{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"name":"python","version":"3.10.12","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":59094,"databundleVersionId":7010844,"sourceType":"competition"}],"dockerImageVersionId":30587,"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# What is about  ?\n\nHere we analyse public vs private scores of our teams submits for 2023 Open problems - \"Single Cell Perturbations\" and 2022 Open problems - \"Multi modal single cell integration\"\n\nCorrelations is very high Person :  0.98 (2023), 0.99 (2022); Spearman: 0.96 (2023), 0.89  (2022).\n\nWelcome to do it for your team for any Kaggle competition:\ngo to the submission page: https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/submissions\nCopy (Ctrl+c) the submission information to buffer.\n\nVersions of the notebook:\n\n    7 Analysis for: 2023 Open problems - \"Single Cell Perturbations\"\n    6 Analysis for: 2022 Open problems - \"Multi modal single cell integration\"\n    \n\nInsert it to the present notebook into the cell below and run the notebook - correltation, scatterplot, lineplot will be created.\n\nThe text copied from Kaggle submissions, page looks like this:\n\n    OP2 some blends - Version 16\n    Complete · Alexander Chervov · 1d ago · blend_558_04netsAntoninaOnFirst128Bcells higher weight 0.4 for Antoninas nets on B-cells\n    0.746\n\n    0.559\n\n    Fork of SCP - blend with class coefs - Version 7\n    Complete · Dmitriy Ershov · 1d ago\n    0.762\n\n    0.571\n\n\nWe copy it as it is , put in triple quotes:  '''text'''' ; and than parse that string.\n\n","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-12-01T20:37:56.032735Z","iopub.execute_input":"2023-12-01T20:37:56.033070Z","iopub.status.idle":"2023-12-01T20:37:56.038870Z","shell.execute_reply.started":"2023-12-01T20:37:56.033046Z","shell.execute_reply":"2023-12-01T20:37:56.037806Z"}}},{"cell_type":"markdown","source":"# The submission info raw text\n\n","metadata":{}},{"cell_type":"code","source":"str_challenge_name = 'SCP (Single-Cell Perturbations)'\nstr_team = 'U900'\n\nif False: # True: # \"True\" that for 2022 Open problems challenge\n    str_challenge_name = 'Multimodal Single-Cell Integration'\n    str_team = 'sB2'\n    \n","metadata":{"execution":{"iopub.status.busy":"2023-12-02T12:25:48.234026Z","iopub.execute_input":"2023-12-02T12:25:48.234449Z","iopub.status.idle":"2023-12-02T12:25:48.241208Z","shell.execute_reply.started":"2023-12-02T12:25:48.234420Z","shell.execute_reply":"2023-12-02T12:25:48.239814Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"str_raw_text = '''                                 \nOP2 some blends - Version 16\nComplete · Alexander Chervov · 1d ago · blend_558_04netsAntoninaOnFirst128Bcells higher weight 0.4 for Antoninas nets on B-cells\n0.746\n\n0.559\n\nFork of SCP - blend with class coefs - Version 7\nComplete · Dmitriy Ershov · 1d ago\n0.762\n\n0.571\n\nOP2 some blends - Version 15\nComplete · Alexander Chervov · 1d ago · blend_558_02netsAntonina 569-571 OnFirst128Bcells\n0.745\n\n0.558\n\nFork of OP2 OOF new folds V5 pyboost no tsvd - Version 11\nComplete · Alexander Chervov · 2d ago · Pyboost train on direct targets - not tsvd for targets, ntree250, 5 random folds, 6maxdepth, targets by 500 in one model\n0.796\n\n0.594\n\nOP2 some blends - Version 13\nComplete · Alexander Chervov · 2d ago · public531_blend_559_02newpyboostAnd587_BoostUp05Mielod\n0.745\n\n0.559\n\nOP2 some blends - Version 12\nComplete · Alexander Chervov · 2d ago · submission_public531_blend_558_ie559and2newpyboostAnd587\n0.744\n\n0.528\n\nOP2 some blends - Version 11\nComplete · Alexander Chervov · 2d ago · blend_559_02newpyboostAnd587\n0.744\n\n0.558\n\nblend_best_on_rem_schemes.csv\nComplete · Antonina Dolgorukova · 2d ago\n0.763\n\n0.568\n\nSCP - pseudo50 ct strat mrrmse tf smilesv - Version 2\nComplete · Dmitriy Ershov · 2d ago\n0.781\n\n0.587\n\nsubmission_0.csv\nComplete · Pizzaboi · 3d ago\n0.804\n\n0.606\n\nOP2 some blends - Version 9\nComplete · Alexander Chervov · 3d ago · NNohe_two_averaged_v106_107 noise0.003 mult20 LR1e-3 BS16 4Layers 256\n0.835\n\n0.619\n\nSCP - blend with class coefs - Version 1\nComplete · Dmitriy Ershov · 3d ago\n0.771\n\n0.591\n\nave_blend_rem9drugs_by_3_3kmeans.csv\nComplete · Antonina Dolgorukova · 3d ago · 9 worst predicted drugs removed by 3 in 15 folds\n0.763\n\n0.573\n\nave_blend_T8_T4_Treg_3kmeans_best_gx3.csv\nComplete · Antonina Dolgorukova · 3d ago · best genes predicted trice with sep features, T8-T4-Treg scheme\n0.766\n\n0.571\n\nSCP - blend own - Version 4\nComplete · Dmitriy Ershov · 4d ago\n0.784\n\n0.59\n\nOP2 some blends - Version 8\nComplete · Alexander Chervov · 4d ago · Tonya566and566NLPandBoosts574_and_575_catboost_pyboost_both584\n0.746\n\n0.559\n\nOP2 some blends - Version 7\nComplete · Alexander Chervov · 4d ago · NLP574_and_575_catboost_pyboost_both584\n0.754\n\n0.566\n\nOP2 some blends - Version 6\nComplete · Alexander Chervov · 4d ago · catboost_and_pyboost_584_average\n0.763\n\n0.575\n\nFork of OP2 OOF new folds V3 - Version 64\nComplete · Alexander Chervov · 4d ago · catboost589, but train on full wo CD8 and 3rand samples excluded 5 times\n0.776\n\n0.584\n\nSCP - blend own - Version 3\nComplete · Dmitriy Ershov · 5d ago\n0.767\n\n0.587\n\nOP2 some blends - Version 5\nComplete · Alexander Chervov · 5d ago · two_catboost5000_md6,7_averaged\n0.783\n\n0.591\n\nave_blend_T8_Treg_3kmeans.csv\nComplete · Antonina Dolgorukova · 5d ago · 16 folds, T8-Treg scheme, 3 groups by kmeans\n0.766\n\n0.57\n\nblend2_corr_cl_plus_scores.csv\nComplete · Antonina Dolgorukova · 5d ago · clusters with same scores averaged, then the lesser score the bigger the weigth\n0.76\n\n0.566\n\nave_blend_T4_NK_minus_T8_3kmeans.csv\nComplete · Antonina Dolgorukova · 6d ago · rem CD8, T4-NK scheme, 3 groups by kmeans\n0.769\n\n0.569\n\nOP2 some blends - Version 4\nComplete · Alexander Chervov · 6d ago · pub547_Tonya_569_570_572_574DimaAlex df5*0.7+ 0.3*(( ( df1 + df2)/2 *0.7 + df3*0.3)*0.7 + df4*0.3)\n0.751\n\n0.542\n\nOP2 some blends - Version 3\nComplete · Alexander Chervov · 6d ago · blend_Tonya_569_570_572_574DimaAlex ( ( df1 + df2)/2 *0.7 + df3*0.3)*0.7 + df4*0.3\n0.751\n\n0.56\n\nOP2 some blends - Version 2\nComplete · Alexander Chervov · 6d ago · blend_Tonya_569_570_572 ( df1 + df2)/2 *0.7 + df3*0.3)\n0.762\n\n0.567\n\nnotebookd29804dc76 - Version 1\nComplete · Alexander Chervov · 6d ago · 574NLPregressor_and_two_more_runs_with_same_prms\n0.762\n\n0.574\n\nSCP - ct stratified mrrmse loss tf smilesvec - Version 1\nComplete · Dmitriy Ershov · 6d ago\n0.787\n\n0.595\n\nNLP_Regression custom_kfold Update1 - Version 120\nComplete · Alexander Chervov · 7d ago · NLPr_SB1_MUL20_Epo5_Noi0.09_RBW1 0.607524 0.436299 0.315357 0.491471 0.314746 0.507401 0.144458 1351\n0.797\n\n0.615\n\nSCP - ct stratifed pt & tf smilesvec - Version 1\nComplete · Dmitriy Ershov · 7d ago\n0.838\n\n0.614\n\nave_blend_T4_T8_3kmeans_sep_f_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 7d ago · blend for T4-T8 scheme, predict 3 groups by kmeans, sep features for each group\n0.764\n\n0.569\n\nNLP_Regression custom_kfold Update1 - Version 103\nComplete · Alexander Chervov · 7d ago · NLPr_SB1_MUL20_Epo20_Noi0.09_RBW0\n0.778\n\n0.598\n\nNLP_Regression custom_kfold Update1 - Version 101\nComplete · Alexander Chervov · 7d ago · NLPr_SB1_MUL20_Epo10_Noi0.09_RBW0\n0.776\n\n0.582\n\nave_blend_Treg_NK_8folds_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 8d ago · 8folds minus 2-3 NK + 8 folds minus 2-3 Treg\n0.765\n\n0.574\n\nave_blend_T4_T8_top10_mark_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 8d ago · like MLPv151 (counts adata TE) on T8-T4 scheme\n0.768\n\n0.576\n\nave_blend_T4_T8_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 8d ago · pseudo labelled 0.571 on T8-T4 scheme\n0.768\n\n0.572\n\nave_blend_0.571_T4_T8_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 8d ago · blend v 0.571 LB with pseudo labelled preds of 0.571 in T4_T8_20ep_augm50_s1_0.1\n0.763\n\n0.57\n\nsubmission_12.5e-05.csv\nComplete · Pizzaboi · 9d ago\n0.866\n\n0.633\n\nNLP_Regression custom_kfold Update1 - Version 44\nComplete · Alexander Chervov · 9d ago · NLPr_SB5_MUL30_Epo300_Noi0.01 C44 0.540 0.396 0.425 0.605 0.427 0.577 0.303 4540.300\n0.811\n\n0.585\n\nNLP_Regression custom_kfold Update1 - Version 5\nComplete · Dmitriy Ershov · 9d ago\n0.889\n\n0.644\n\nave_all_train_614_20ep_augm50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 9d ago · 16 folds trained on the entire train\n0.767\n\n0.574\n\nsubmission_1.09.csv\nComplete · Pizzaboi · 10d ago\n0.867\n\n0.642\n\nOP2: blend/ensemble - Version 58\nComplete · Antonina Dolgorukova · 10d ago · 8 folds minus 2-3 CD4, 8 folds minus 2-3 CD8\n0.763\n\n0.571\n\nsubmit_test_cells_61folds_augm50_s1_0.1_mrRMSE_0.983.csv\nComplete · Antonina Dolgorukova · 10d ago · 61 fold, each without 10-14 obs.\n0.763\n\n0.574\n\nOP2: blend/ensemble - Version 57\nComplete · Antonina Dolgorukova · 10d ago · conc_blend_v49_0.574 - same models averaged\n0.761\n\n0.572\n\nNLP_Regression custom_kfold Update1 - Version 11\nComplete · Alexander Chervov · 10d ago · Multiplex25train NLP_reg Tonya-CV-scores: 0.569 0.414 0.377 0.564 0.365 0.544 0.236 3105.100 C11\n0.766\n\n0.574\n\nMultiseed NLP regression - Version 2\nComplete · Pizzaboi · 11d ago · 25 different seeds\n0.786\n\n0.586\n\nOP2: blend/ensemble - Version 56\nComplete · Antonina Dolgorukova · 11d ago · V128 8folds minus 2-3 Myel + 8 folds minus 2-3 B (concat)\n0.766\n\n0.578\n\nOP2: blend/ensemble - Version 52\nComplete · Antonina Dolgorukova · 11d ago · as blend v49 - Myel20 + B cells 50 (8 folds in each) but without CD8\n0.762\n\n0.576\n\nOP2: blend/ensemble - Version 51\nComplete · Antonina Dolgorukova · 11d ago · 9 models with max diversity - averaged\n0.772\n\n0.573\n\nsubmit_603_ave_next_augm50_s1_0.1_mrRMSE_0.767.csv\nComplete · Antonina Dolgorukova · 11d ago\n0.783\n\n0.582\n\nNLP_Regression Pseudolabeling - Version 3\nComplete · Pizzaboi · 12d ago · 20% pseudo\n0.784\n\n0.585\n\nNLP_Regression Pseudolabeling - Version 2\nComplete · Pizzaboi · 12d ago\n0.783\n\n0.589\n\nNLP_Regression Pseudolabeling - Version 1\nComplete · Pizzaboi · 12d ago\n0.782\n\n0.583\n\nSCP - Pytorch & Keras NN cl+te+tsne - Version 1\nComplete · Dmitriy Ershov · 14d ago\n0.896\n\n0.662\n\nsubmission_3kmgr_aug50_s1_0.1.csv\nComplete · Antonina Dolgorukova · 15d ago\n0.797\n\n0.582\n\nsubmission_3kmgr_aug50_s1_0.1.csv\nError · Antonina Dolgorukova · 15d ago · 3L5f CV on test drugs, pcaTE augm 50, s1/0.1, targ TE features 100, 100, km 3gr\nFork of Fork of NLP_Regression 12a31a 2fe170 - Version 4\nComplete · Pizzaboi · 15d ago\n0.776\n\n0.581\n\nFork of Fork of NLP_Regression 12a31a 2fe170 - Version 2\nComplete · Pizzaboi · 15d ago\n0.778\n\n0.582\n\nFork of Fork of NLP_Regression 12a31a 2fe170 - Version 1\nComplete · Pizzaboi · 15d ago\n0.777\n\n0.584\n\nsubmission_noise_to_test_f.csv\nComplete · Antonina Dolgorukova · 16d ago · augm test features * 50 + noise 1\n0.79\n\n0.583\n\nOP2: MLP submission - Version 207\nComplete · Antonina Dolgorukova · 16d ago · scaled_counts_all_genes 20/20 + TEmean 80/80 augm 50, s1/1\n0.804\n\n0.586\n\nOP2: MLP submission - Version 208\nComplete · Antonina Dolgorukova · 16d ago · as v49 607 cells\n0.796\n\n0.58\n\nOP2: blend/ensemble - Version 50\nComplete · Antonina Dolgorukova · 16d ago · as v49 (myel + b) but on data from > 2 n_cells\n0.784\n\n0.581\n\nSCP - Pytorch & Keras NN Clusters data - Version 3\nComplete · Dmitriy Ershov · 16d ago\n0.902\n\n0.666\n\nOP2: blend/ensemble - Version 49\nComplete · Antonina Dolgorukova · 17d ago · 8 folds minus 2-3 Myel + 8 folds minus 2-3 B (concat)\n0.763\n\n0.574\n\nFork of NLP_Regression 12a31a 2fe170 - Version 19\nComplete · Pizzaboi · 17d ago\n0.766\n\n0.587\n\nOP2 blend simple - Version 5\nComplete · Alexander Chervov · 17d ago · [:128]= 07*566+0.3*574pyb, [128:] = 0.4 , 0.6\n0.758\n\n0.562\n\nOP2 blend simple - Version 3\nComplete · Alexander Chervov · 17d ago · 3 give pyboost574 larger weight on [128:] , blend pyboost 574 with our 566\n0.757\n\n0.563\n\nOP2: sep MLPs submission - Version 29\nComplete · Antonina Dolgorukova · 18d ago · as v49 but predict targets PCA and inverse\n0.862\n\n0.631\n\nOP2: MLP submission - Version 184\nComplete · Antonina Dolgorukova · 18d ago · test cells*4 + aug 50. all as in v49\n0.806\n\n0.593\n\nFork of Fork of NLP_Regression BN - Version 3\nComplete · Pizzaboi · 18d ago\n0.79\n\n0.596\n\nFork of NLP_Regression 12a31a 2fe170 - Version 18\nComplete · Pizzaboi · 18d ago\n0.772\n\n0.584\n\nOP2 blend simple - Version 2\nComplete · Alexander Chervov · 18d ago · submission_06_566_04_574publicPyboost.csv\n0.756\n\n0.563\n\nOP2: blend/ensemble - Version 47\nComplete · Antonina Dolgorukova · 18d ago · 6 MLPs from target + raw counts TE (replics + diff feature subsets)\n0.787\n\n0.58\n\nOP2: MLP submission - Version 181\nComplete · Antonina Dolgorukova · 18d ago · TE targ + raw counts, all by 50, *50, s1/0\n0.791\n\n0.585\n\nFork of NLP_Regression 12a31a 2fe170 - Version 15\nComplete · Pizzaboi · 18d ago\n0.79\n\n0.595\n\nOP2: sep MLPs submission - Version 19\nComplete · Antonina Dolgorukova · 18d ago · 13random groups, pcaTE augm 50, s0.1/0.1, 20/20 features\n0.795\n\n0.587\n\nFork of NLP_Regression 12a31a 2fe170 - Version 13\nComplete · Pizzaboi · 18d ago\n0.831\n\n0.633\n\nOP2: MLP submission - Version 170\nComplete · Antonina Dolgorukova · 18d ago · as v49 augm 150\n0.786\n\n0.579\n\nFork of NLP_Regression 12a31a 2fe170 - Version 8\nComplete · Pizzaboi · 18d ago\n0.792\n\n0.592\n\nFork of NLP_Regression 12a31a 2fe170 - Version 7\nComplete · Pizzaboi · 19d ago\n0.781\n\n0.581\n\nFork of NLP_Regression 12a31a 2fe170 - Version 6\nComplete · Pizzaboi · 19d ago\n0.849\n\n0.63\n\nKfold simple NN - Version 1\nComplete · Pizzaboi · 19d ago\n0.875\n\n0.645\n\nFork of NLP_Regression 12a31a 2fe170 - Version 5\nComplete · Pizzaboi · 19d ago\n0.841\n\n0.628\n\nFork of NLP_Regression 12a31a 2fe170 - Version 3\nComplete · Pizzaboi · 19d ago\n0.777\n\n0.587\n\nFork of NLP_Regression 12a31a 2fe170 - Version 2\nComplete · Pizzaboi · 19d ago\n0.891\n\n0.659\n\nFork of NLP_Regression 12a31a 2fe170 - Version 1\nComplete · Pizzaboi · 19d ago\n0.789\n\n0.586\n\nOP2 \"Gentle\" param tuner - Version 61\nComplete · Alexander Chervov · 19d ago · Quantile0.8 CatB submission_tsvd30_CATB_QuantileEncoderCompoundCellTypeRandomRandom_tuned.csv\n0.774\n\n0.589\n\nOP2: MLP submission - Version 164\nComplete · Antonina Dolgorukova · 19d ago · rep 49\n0.789\n\n0.583\n\nOP2: MLP submission - Version 154\nComplete · Antonina Dolgorukova · 19d ago · rep 49\n0.792\n\n0.581\n\nOP2: blend/ensemble - Version 46\nComplete · Antonina Dolgorukova · 19d ago · v49 - average of 4 repeats\n0.772\n\n0.577\n\n【0.584】Feature Augmentation - LightGBM - Version 1\nComplete · Dmitriy Ershov · 19d ago\n0.776\n\n0.584\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 1\nComplete · Dmitriy Ershov · 19d ago\n0.859\n\n0.641\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 1\nComplete · Dmitriy Ershov · 19d ago\n0.786\n\n0.602\n\nOP2: MLP submission - Version 158\nComplete · Antonina Dolgorukova · 19d ago · sum raw counts features *50, s1/0.01\n0.899\n\n0.65\n\nOP2 \"Gentle\" param tuner - Version 50\nComplete · Alexander Chervov · 20d ago · 1.180979 CV Random submission_tsvd30_CATB_LeaveOneOutEncoderCompoundCellTypeRandomRandom_tuned.csv\n0.814\n\n0.62\n\nOP2 \"Gentle\" param tuner - Version 52\nComplete · Alexander Chervov · 20d ago · submission_tsvd30_CATB_LeaveOneOutEncoderCompoundCellTypeRandomRandom_tuned.csv\n0.789\n\n0.607\n\nOP2 \"Gentle\" param tuner - Version 52\nComplete · Alexander Chervov · 20d ago · submission_tsvd30_CATB_LeaveOneOutEncoderCompoundCellTypeMTMT_tuned.csv\n0.789\n\n0.607\n\nOP2 \"Gentle\" param tuner - Version 52\nComplete · Alexander Chervov · 20d ago · submission_tsvd30_CATB_LeaveOneOutEncoderCompoundCellTypeAmbrosMAmbrosM_tuned.csv\n0.799\n\n0.614\n\nOP2 blend simple - Version 1\nComplete · Alexander Chervov · 20d ago · 06_572_04_577public\n0.761\n\n0.566\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 3\nComplete · Pizzaboi · 20d ago\n0.789\n\n0.604\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 3\nError · Pizzaboi · 20d ago\nPYBOOST - \"secret\" Grandmaster's tool - Version 2\nComplete · Pizzaboi · 20d ago\n0.79\n\n0.604\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 1\nComplete · Pizzaboi · 20d ago\n0.792\n\n0.605\n\nPYBOOST - \"secret\" Grandmaster's tool - Version 1\nError · Pizzaboi · 20d ago\nOP2: blend/ensemble - Version 45\nComplete · Antonina Dolgorukova · 21d ago · blend on logic (10 diverse MLPs + 27 good folds ), this * 0.95 + MLPv151 0.583 * 0.05, this * 0.8 + MLPv80 0.59 * 0.2\n0.768\n\n0.574\n\nOP2: from PC submit + ensemble - Version 44\nComplete · Antonina Dolgorukova · 21d ago · Logic cluster 1, 10 models\n0.772\n\n0.574\n\nOP2: from PC submit + ensemble - Version 41\nComplete · Antonina Dolgorukova · 22d ago · cluster MLP on rand good folds: averaged 5*0.85 + MPLv58_0.603 * 0.15\n0.768\n\n0.578\n\nOP2: from PC submit + ensemble - Version 40\nComplete · Antonina Dolgorukova · 22d ago · blend v39 * 0.95 + v58_0.603 *0.05\n0.77\n\n0.572\n\nOP2: from PC submit + ensemble - Version 39\nComplete · Antonina Dolgorukova · 22d ago · ( (cl1 + cl3) / 2 ) *0.8 + ( (cl2 + v105) / 2 ) * 0,2 это с коэффом 0,8 + v58 * 0.2\n0.769\n\n0.572\n\nOP2: from PC submit + ensemble - Version 38\nComplete · Antonina Dolgorukova · 22d ago · cl 3 16 models\n0.767\n\n0.576\n\nOP2: from PC submit + ensemble - Version 37\nComplete · Antonina Dolgorukova · 22d ago · cl 2 (MLPv32_0.582 + MLPv35_0.582) /2\n0.764\n\n0.58\n\nOP2: from PC submit + ensemble - Version 36\nComplete · Antonina Dolgorukova · 22d ago · cl 1 (MLPv122_0.581 + MLPv127_0.583 + MLPv151_0.583) /3\n0.789\n\n0.576\n\nOP2: MLP submission - Version 151\nComplete · Antonina Dolgorukova · 22d ago · pcaTE 80/80 + count_sums_features TE 20/20\n0.793\n\n0.583\n\nOP2 pyboost - Version 19\nComplete · Alexander Chervov · 22d ago · Pyboost_max_depth10_ntrees1000_lr001_subsample1_colsample02_n_components50\n0.79\n\n0.604\n\nOP2: MLP submission - Version 149\nComplete · Antonina Dolgorukova · 23d ago · noise to targets 0.1\n0.788\n\n0.581\n\nOP2: MLP submission - Version 145\nComplete · Antonina Dolgorukova · 23d ago · as v128, 20% aug by rows\n0.791\n\n0.582\n\nOP2: MLP submission - Version 128\nComplete · Antonina Dolgorukova · 23d ago · pcaTE augm 30, + by rows 10%, 0.5/0.5\n0.788\n\n0.582\n\nOP2: MLP submission - Version 138\nComplete · Antonina Dolgorukova · 23d ago · pcaTE +augm 50 s1/0.1 + by row 10%, 0.5/0.5\n0.794\n\n0.588\n\nOP2: MLP submission - Version 127\nComplete · Antonina Dolgorukova · 24d ago · *10, 10%, 0.5/0.5\n0.791\n\n0.583\n\nOP2: MLP submission - Version 125\nComplete · Antonina Dolgorukova · 24d ago · 0.1 augm by row 0.1/0.9, targets PCA TE,\n0.822\n\n0.603\n\nOP2: MLP submission - Version 112\nComplete · Antonina Dolgorukova · 25d ago · s 1 2cells filt before PCA\n0.786\n\n0.582\n\nOP2: MLP submission - Version 122\nComplete · Antonina Dolgorukova · 25d ago · TE augm 50, s1/0.1, n_cells > 5 filt bedore PCA\n0.802\n\n0.581\n\nOP2: MLP submission - Version 121\nComplete · Antonina Dolgorukova · 25d ago · sigma 1/0, ncells filt before PCA\n0.784\n\n0.58\n\nOP2: glmnet drug+cell submission - Version 25\nComplete · Antonina Dolgorukova · 25d ago · sep model with 2 features - targets\n1.035\n\n0.737\n\nOP2: MLP submission - Version 111\nComplete · Antonina Dolgorukova · 25d ago · Raw counts features PCA TE, 20/20, no augm, s 0.01\n0.902\n\n0.664\n\nOP2: MLP submission - Version 108\nComplete · Antonina Dolgorukova · 25d ago · Raw counts features PCA TE\n1.069\n\n0.773\n\nOP2 blend with JAX - Version 1\nComplete · Alexander Chervov · 25d ago · blend01JaxWithLB570Antonina\n0.758\n\n0.571\n\nOP2: from PC submit + ensemble - Version 31\nComplete · Antonina Dolgorukova · 1mo ago · (MLPv12_0.606 + pubNNv4_0.609 + op2v8_0.614 + Random_20_42 + Random_20_42) / 5 * 0.15 + blend 0.570 * 0.85\n0.763\n\n0.571\n\nOP2: from PC submit + ensemble - Version 30\nComplete · Antonina Dolgorukova · 1mo ago · (MLPv12_0.606 + pubNNv4_0.609 + op2v8_0.614 + Random_20_42 + Random_20_42) / 5 *0.85 + blend 0.570 * 0.15\n0.79\n\n0.59\n\nOP2: from PC submit + ensemble - Version 29\nComplete · Antonina Dolgorukova · 1mo ago · (MLPv12_0.606 + pubNNv4_0.609 + op2v8_0.614 + Random_20_42 + Random_20_42) / 5\n0.799\n\n0.597\n\nOP2: MLP submission - Version 105\nComplete · Antonina Dolgorukova · 1mo ago · as v49 60/60 features\n0.766\n\n0.583\n\nOP2: MLP submission - Version 107\nComplete · Antonina Dolgorukova · 1mo ago · as v49, 140/100 features\n0.771\n\n0.581\n\nOP2: MLP submission - Version 99\nComplete · Antonina Dolgorukova · 1mo ago · features_tsvd100_encoded_LeaveOneOutEncoder\n1.373\n\n0.939\n\nOP2: MLP submission - Version 86\nComplete · Antonina Dolgorukova · 1mo ago · as v49 but rrelu, pt = 5\n0.766\n\n0.582\n\nOP2: from PC submit + ensemble - Version 28\nComplete · Antonina Dolgorukova · 1mo ago · 0.571 * 0.8 + MLPv80_0.59 * 0.2\n0.76\n\n0.57\n\nOP2: from PC submit + ensemble - Version 26\nComplete · Antonina Dolgorukova · 1mo ago · mean( 0.571, MLPv80_0.59 )\n0.765\n\n0.573\n\nOP2 correlate and submit - Version 2\nComplete · Alexander Chervov · 1mo ago · submission_571blendwithTwo608AlexWeights0802\n0.764\n\n0.573\n\nOP2 correlate and submit - Version 1\nComplete · Alexander Chervov · 1mo ago · 571blendwithTwo608Alex (CatB and SVR TE i-th-only )\n0.775\n\n0.581\n\nOP2: MLP submission - Version 78\nComplete · Antonina Dolgorukova · 1mo ago · as v49, 5 rand folds, thr 0.8, split 0.8\n0.769\n\n0.582\n\nOP2: MLP submission - Version 77\nComplete · Antonina Dolgorukova · 1mo ago · as v49, 5 rand folds, thr 0.8, split 0.85\n0.77\n\n0.582\n\nOP2: MLP submission - Version 82\nComplete · Antonina Dolgorukova · 1mo ago · as v12 but ICA100\n0.844\n\n0.627\n\nOP2: MLP submission - Version 81\nComplete · Antonina Dolgorukova · 1mo ago · as v12 but tsvd\n0.803\n\n0.61\n\nOP2: MLP submission - Version 80\nComplete · Antonina Dolgorukova · 1mo ago · as v49, but tsvd\n0.784\n\n0.59\n\nOP2 blends - Version 6\nComplete · Alexander Chervov · 1mo ago · blend 0.574 Antonina 0.583 public 0.5, 0.5\n0.764\n\n0.572\n\nOP2 blends - Version 5\nComplete · Alexander Chervov · 1mo ago · blend 0.574 Antonina 0.583 public 0.7, 0.3\n0.761\n\n0.571\n\nOP2 blends - Version 4\nComplete · Alexander Chervov · 1mo ago · blend 0.574 Antonina 0.583 public 0.85, 0.15\n0.761\n\n0.571\n\nOP2 \"Gentle\" param tuner - Version 2\nComplete · Alexander Chervov · 1mo ago · tsvd30_KRRlin_LeaveOneOutEncoderCompound optmized Random CV 1.2059 blended with priors 0.45, 0.1\n0.813\n\n0.605\n\nOP2 \"Gentle\" param tuner - Version 2\nComplete · Alexander Chervov · 1mo ago · tsvd30_KRRlin_LeaveOneOutEncoderCompound 0.959825 Optimized AmbrosM blended with priors\n0.814\n\n0.605\n\nOP2 \"Gentle\" param tuner - Version 2\nComplete · Alexander Chervov · 1mo ago · tsvd30_KRRlin_LeaveOneOutEncoderCompound MT-optimized CV: 2.535224 blend with priors 0.45,0.1\n0.811\n\n0.613\n\nOP2 \"Gentle\" param tuner - Version 1\nError · Alexander Chervov · 1mo ago · KRR-linear tsvd30 compound LOO 2.535224 MT-CV blend with priors 0.45 0.1\nOP2: MLP submission - Version 66\nComplete · Antonina Dolgorukova · 1mo ago · v49, Random CV with threshold 1.5\n0.775\n\n0.583\n\nOP2 target encoders - Version 3\nComplete · Alexander Chervov · 1mo ago · LOO-Encoder Compound only tsvd30 Ridge Blend with priors\n0.815\n\n0.608\n\nOP2: MLP submission - Version 59\nComplete · Antonina Dolgorukova · 1mo ago · TE + pca features (all 152)\n0.835\n\n0.625\n\nOP2 category encoders, chembert,fingerpints,moldes - Version 27\nComplete · Alexander Chervov · 1mo ago · RidgeAlpha15000_HelmertEncoderCompoundCellType_tsvd30_blendWithPriors_0.45_0.1\n0.819\n\n0.605\n\nOP2: from PC submit + ensemble - Version 23\nComplete · Antonina Dolgorukova · 1mo ago · 9 averaged MLP*0.9 + 2good folds*0.1 MLP_v48_0.58.csv MLPv32_0.582.csv MLPv35_0.582.csv MLPv39_0.581.csv MLPv46_0.581.csv MLPv49_0.579.csv MLPv54_0.584.csv MLPv56_0.585.csv MLPv57_0.581.csv\n0.763\n\n0.574\n\nOP2: from PC submit + ensemble - Version 22\nComplete · Antonina Dolgorukova · 1mo ago · mean(MLPv54_0.584 + MLPv56_0.585 + MLPv57_0.581)\n0.766\n\n0.58\n\nOP2: from PC submit + ensemble - Version 21\nComplete · Antonina Dolgorukova · 1mo ago · mean(MLPv54_0.584 + MLPv56_0.585 + MLPv57_0.581) * 0.75 + MPLv58_0.603 * 0.25\n0.771\n\n0.577\n\nOP2: MLP submission - Version 58\nComplete · Antonina Dolgorukova · 1mo ago · 2 rand of 20 l, threshold 0.9\n0.824\n\n0.603\n\nOP2 testing combines - Version 6\nComplete · Alexander Chervov · 1mo ago · combined 01\n0.788\n\n0.589\n\nOP2 testing combines - Version 5\nComplete · Alexander Chervov · 1mo ago · combined10\n0.789\n\n0.589\n\nOP2 testing combines - Version 4\nComplete · Alexander Chervov · 1mo ago · comb08\n0.789\n\n0.589\n\nOP2: MLP submission - Version 57\nComplete · Antonina Dolgorukova · 1mo ago · rand 5 folds/10, threshold 1\n0.763\n\n0.581\n\nOP2 testing combines - Version 3\nComplete · Alexander Chervov · 1mo ago · combine2\n0.788\n\n0.589\n\nOP2 class for custom CV schemes - Version 8\nComplete · Alexander Chervov · 1mo ago · RidgeAlpha1 onehotDrug tsvd30 blend priors 0.35 0.1\n0.814\n\n0.605\n\nOP2: MLP submission - Version 56\nComplete · Antonina Dolgorukova · 1mo ago · rand 5 folds, threshold 1.1\n0.775\n\n0.585\n\nOP2: MLP submission - Version 54\nComplete · Antonina Dolgorukova · 1mo ago · 5 rand folds, thr 1.2\n0.772\n\n0.584\n\nOP2: MLP submission - Version 52\nComplete · Antonina Dolgorukova · 1mo ago · rep v49\n0.768\n\n0.58\n\nOP2: from PC submit + ensemble - Version 20\nComplete · Antonina Dolgorukova · 1mo ago · mse_loss, drop 0.2, drop 0.1\n0.82\n\n0.602\n\nOP2: from PC submit + ensemble - Version 19\nComplete · Antonina Dolgorukova · 1mo ago · v51 repeat\n0.786\n\n0.597\n\nOP2: MLP submission - Version 51\nComplete · Antonina Dolgorukova · 1mo ago · s1, mse_loss\n0.78\n\n0.59\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 469\nComplete · Alexander Chervov · 1mo ago · combined09 0.07 0.01\n0.789\n\n0.589\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 469\nComplete · Alexander Chervov · 1mo ago · combined09 0.07 0.01\n1.496\n\n1\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 468\nComplete · Alexander Chervov · 1mo ago · combined09 0.03 0.01\n0.789\n\n0.589\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 466\nComplete · Alexander Chervov · 1mo ago · 466 0.97 combined09 0.02 0.01\n0.789\n\n0.589\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 463\nComplete · Alexander Chervov · 1mo ago · combined 09top1 + 01 615\n0.79\n\n0.59\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 462\nComplete · Alexander Chervov · 1mo ago · 04_0608_00_0610_614_06_615.\n0.805\n\n0.602\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 461\nComplete · Alexander Chervov · 1mo ago · 05_0608_00_0610_614_05_615\n0.804\n\n0.601\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 460\nComplete · Alexander Chervov · 1mo ago · 06_0608_00_0610_614_04_615\n0.803\n\n0.601\n\nOP2 Kishan's NN streamlined and blended - Version 2\nComplete · Alexander Chervov · 1mo ago · 100 self blends Kishans NN\n0.831\n\n0.621\n\nOP2 Kishan's NN streamlined and blended - Version 6\nComplete · Alexander Chervov · 1mo ago · 2000 self blends Kishan NN\n0.828\n\n0.62\n\nOP2: MLP submission - Version 50\nComplete · Antonina Dolgorukova · 1mo ago · S 1.5\n0.763\n\n0.58\n\nOP2: from PC submit + ensemble - Version 18\nComplete · Antonina Dolgorukova · 1mo ago\n0.761\n\n0.575\n\nOP2: from PC submit + ensemble - Version 17\nComplete · Antonina Dolgorukova · 1mo ago\n0.76\n\n0.575\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 458\nComplete · Alexander Chervov · 1mo ago · 458 0.6*608+0.1*610-614+0.3*615\n0.804\n\n0.601\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 457\nComplete · Alexander Chervov · 1mo ago · 457 0.7*608 + 0.1*610-614 + 0.2*615\n0.803\n\n0.601\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 456\nComplete · Alexander Chervov · 1mo ago · 07_0608_02_0610_614_01_615\n0.804\n\n0.602\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 455\nComplete · Alexander Chervov · 1mo ago · 08_0608_01_0610_614_01_615\n0.803\n\n0.602\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 455\nComplete · Alexander Chervov · 1mo ago · 08_0608_01_0610_614_01_615 + 02Drug_01CellType\n0.815\n\n0.606\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 402\nComplete · Alexander Chervov · 1mo ago · 0.9656 2.8382 op2cv 402 1.182254 1h46min krr rbf tsvd25_KRR_QuantileEncoder_BothEncoded\n0.841\n\n0.619\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 407\nComplete · Alexander Chervov · 1mo ago · 0.951196 2.967457 opAm 407 1.199421 1h7min svr-rbf\n0.848\n\n0.626\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 408\nComplete · Alexander Chervov · 1mo ago · 0.962 2.746386 op2Am 408 1.19523 svr-linear tsvd25_SVR_QuantileEncoder_BothEncoded\n0.843\n\n0.619\n\nOP2: from PC submit + ensemble - Version 16\nComplete · Antonina Dolgorukova · 1mo ago\n0.762\n\n0.578\n\nOP2: from PC submit + ensemble - Version 15\nComplete · Antonina Dolgorukova · 1mo ago\n0.761\n\n0.575\n\nOP2: from PC submit + ensemble - Version 14\nComplete · Antonina Dolgorukova · 1mo ago\n0.761\n\n0.575\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 405\nComplete · Alexander Chervov · 1mo ago · 0.9657 2.6849 svr rbf op2 405 1.197947 1h12min svr rbf tsvd25_SVR_QuantileEncoder_BothEncoded\n0.822\n\n0.619\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 394\nComplete · Alexander Chervov · 1mo ago · 0.972 2.67 Quartile Ridge one-prm-all\n0.91\n\n0.657\n\nOP2: MLP submission - Version 49\nComplete · Antonina Dolgorukova · 1mo ago · sigma 1\n0.764\n\n0.579\n\nOP2: MLP submission - Version 48\nComplete · Antonina Dolgorukova · 1mo ago · sigma 0.5\n0.766\n\n0.58\n\nOP2: MLP submission - Version 47\nComplete · Antonina Dolgorukova · 1mo ago · sigma 0.05\n0.775\n\n0.587\n\nOP2: from PC submit + ensemble - Version 13\nComplete · Antonina Dolgorukova · 1mo ago\n0.762\n\n0.575\n\nOP2: from PC submit + ensemble - Version 12\nComplete · Antonina Dolgorukova · 1mo ago\n0.764\n\n0.576\n\nOP2: from PC submit + ensemble - Version 11\nComplete · Antonina Dolgorukova · 1mo ago · MLPv46_0.581 *0.45 + mean by drug * 0.45 + mean by ct * 0.1\n0.785\n\n0.585\n\nOP2: from PC submit + ensemble - Version 10\nComplete · Antonina Dolgorukova · 1mo ago · mean(MLPv46_0.581 + MLPv39_0.581)\n0.768\n\n0.579\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 329\nComplete · Alexander Chervov · 1mo ago · 1.0099 2.5706 329KRR,ChemBert-cls,Opt2CV-all-tgs,tsvd50\n0.826\n\n0.623\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 368\nComplete · Alexander Chervov · 1mo ago · 0.9888-0.9900, 2.66748 RFR&TE-ith, tsvd25,OptAmbr, 1prm-all ( Reproduce 258 - not saved - simple, not bad\n0.869\n\n0.653\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 353\nComplete · Alexander Chervov · 1mo ago · 0.9787 2.5917 353KRR-rbf,Opt2-all-tgs,Helm-encBoth,tsvd30\n0.822\n\n0.618\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 358\nComplete · Alexander Chervov · 1mo ago · LB0.625 0.9985 2.609 358CatB,Opt2-all-tgs,Helm-encDr,tsvd30 # 7h48min # Not bad LB and CV , but not so good\n0.836\n\n0.625\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 358\nComplete · Alexander Chervov · 1mo ago · LB0.625 0.9985 2.609 358CatB,Opt2-all-tgs,Helm-encDr,tsvd30 # 7h48min # Not bad LB and CV , but not so good\n0.835\n\n0.624\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 254\nComplete · Alexander Chervov · 1mo ago · 0.9683 2.4244 254CatB,TE-ith,Opt2-all-tgs&SM,tsvd25,1R, #5h\n0.801\n\n0.608\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 255\nComplete · Alexander Chervov · 1mo ago · 0.956 2.597715 255CatB,TE-ith,OptAm-all-tgs&SM,tsvd25,1R, #3.5h #\n0.832\n\n0.633\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 243\nComplete · Alexander Chervov · 1mo ago · 0.9710 2.4709 243Catb,TE-ith,OptAm-all-tgs,tsvd25,1R\n0.837\n\n0.621\n\nOP2: from PC submit + ensemble - Version 9\nComplete · Antonina Dolgorukova · 1mo ago · mean(MLPv46_0.581 + MLPv39_0.581)*0.45 + mean by drug * 0.45 + mean by ct * 0.1\n0.785\n\n0.585\n\nOP2: from PC submit + ensemble - Version 8\nComplete · Antonina Dolgorukova · 1mo ago · PCA > 2 PC inverse PCA\n0.787\n\n0.596\n\nOP2: from PC submit + ensemble - Version 7\nComplete · Antonina Dolgorukova · 1mo ago · v46 pca 20 PC > inverse\n0.772\n\n0.581\n\nOP2: MLP submission - Version 46\nComplete · Antonina Dolgorukova · 1mo ago · sigma 0.3\n0.772\n\n0.581\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 255\nComplete · Alexander Chervov · 1mo ago · 0.956 2.597715 255CatB,TE-ith,OptAm-all-tgs&SM,tsvd25,1R, #3.5h\n0.832\n\n0.633\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 259\nComplete · Alexander Chervov · 1mo ago · # 0.9939 2.5496 RFR 1prm-all - huge random variation: 0.985899 - 2.580414 259 rfr op2cv # huge random variation from run to run 'n_estimators': 30, 'max_depth': 2, 'reducer':\n0.913\n\n0.689\n\nOP2: MLP submission - Version 45\nComplete · Antonina Dolgorukova · 1mo ago · 170 f pc\n0.782\n\n0.588\n\nOP2: MLP submission - Version 43\nComplete · Antonina Dolgorukova · 1mo ago · TE +augm 50, n_cells >2 pairs, - controls\n0.786\n\n0.607\n\nOP2: MLP submission - Version 42\nComplete · Antonina Dolgorukova · 1mo ago · >5\n0.799\n\n0.583\n\nOP2: MLP submission - Version 41\nComplete · Antonina Dolgorukova · 1mo ago · n_cells >10 pairs\n0.814\n\n0.598\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 242\nComplete · Alexander Chervov · 1mo ago · 1.07448 2.3448 242Catb,TE-ith,OptMT-all-tgs,tsvd25,1R\n0.838\n\n0.635\n\nOP2: MLP submission - Version 40\nComplete · Antonina Dolgorukova · 1mo ago · v39 lr 0.05\n0.789\n\n0.589\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 244\nComplete · Alexander Chervov · 1mo ago · 0.9779 2.4194 244Catb,TE-ith,Opt2-all-tgs,tsvd25,1R\n0.817\n\n0.612\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 246\nComplete · Alexander Chervov · 1mo ago · 1.0009 2.6823 246LGB,Opt2-all-tgs,te-ith,tsvd25,2R\n0.823\n\n0.616\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 240\nComplete · Alexander Chervov · 1mo ago · 0.9908 2.517472 240Catb,TE-ith,OptAm-1prm-all,tsvd25,2R #9min # 'iterations': 10, 'depth': 3, 'learning_rate': 0.2,\n0.816\n\n0.627\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 237\nComplete · Alexander Chervov · 1mo ago · 1.0027 2.6498 237CatB,oheDr,tsvd25,OptAmMT-all-tgs,1R\n0.83\n\n0.622\n\nOP2: MLP submission - Version 38\nComplete · Antonina Dolgorukova · 1mo ago · *50, tuned\n0.763\n\n0.599\n\nOP2: MLP submission - Version 39\nComplete · Antonina Dolgorukova · 1mo ago\n0.771\n\n0.581\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 222\nComplete · Alexander Chervov · 1mo ago · 0.9946 2.6139 222tsvd50,oheDrg,KRR-rbf,Opt2-all-tgs,Al+ # 6h 0.01,0.009,0.008,0.005,0.011,0.015,0.02, 0.001,0.0001,0.1,1,10\n0.822\n\n0.615\n\nOP2: MLP submission - Version 37\nComplete · Antonina Dolgorukova · 1mo ago · *25\n0.768\n\n0.586\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 210\nComplete · Alexander Chervov · 1mo ago · 1.003 2.6579 'alpha': 0.01 210oheDrg,KRR-rbf,tsvd25, Opt2-1prm-4-all,AlphaExt\n0.823\n\n0.614\n\nOP2: MLP submission - Version 36\nComplete · Antonina Dolgorukova · 1mo ago · v32 seed 5\n0.767\n\n0.586\n\nOP2: MLP submission - Version 35\nComplete · Antonina Dolgorukova · 1mo ago · *100\n0.767\n\n0.582\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 202\nComplete · Alexander Chervov · 1mo ago · Randomized no priors svr-l , te-ith , tsvd50 opr2cv-all-targets c, sm, 0.608\n0.819\n\n0.607\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 202\nComplete · Alexander Chervov · 1mo ago · With prior 0.608 svr-l tsvd50 te-ith, opt2cv c, smooth\n0.819\n\n0.607\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 202\nComplete · Alexander Chervov · 1mo ago · Svr-l tsvd50 op2all-tgs C , Smooth, Te-ith Rnd1\n0.82\n\n0.608\n\nOP2: MLP submission - Version 33\nComplete · Antonina Dolgorukova · 1mo ago · *50, train_loss\n0.78\n\n0.583\n\nOP2: MLP submission - Version 32\nComplete · Antonina Dolgorukova · 1mo ago · *50 valid_loss\n0.767\n\n0.582\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 145\nComplete · Alexander Chervov · 2mo ago · # 0.9721 2.6752 145 OptAm one-for-all lin-SVR tsvd30 TE-ithBoth # 9h\n0.823\n\n0.61\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 151\nComplete · Alexander Chervov · 2mo ago · # 0.9812 2.6681 151 Opt2 one-for-all lin-SVR tsvd30 TE-ithBoth # 7h18m\n0.832\n\n0.615\n\nOP2: MLP submission - Version 31\nComplete · Antonina Dolgorukova · 2mo ago · *10\n0.772\n\n0.595\n\nOP2: MLP submission - Version 30\nComplete · Antonina Dolgorukova · 2mo ago · v12, augm train * 5, sigma 0.1\n0.778\n\n0.603\n\nOP2: MLP submission - Version 29\nComplete · Antonina Dolgorukova · 2mo ago · v12, augm train * 3, sigma 0.1\n0.785\n\n0.604\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 117\nComplete · Alexander Chervov · 2mo ago · 0.9919 2.6218 Ridge onehot drug tsvd30 117 45m Opt2CV\n0.824\n\n0.616\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 118\nComplete · Alexander Chervov · 2mo ago · 0.9872 2.631839 Ridge onehot drug tsvd30 118 33m OptAmbrosM\n0.824\n\n0.616\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 112\nComplete · Alexander Chervov · 2mo ago · 0.9935 2.6064 Ridge TE-i-th-only tsvd100 112 5h18m Opt i_targ>50 Opt2CV dict_prms_lists = { 'smoothing0': [0,1e4, 1e1,1e2] , 'smoothing1': [1e4, 1e1,1e2,0], 'alpha': [1e4,1e3,1e2],}\n0.814\n\n0.61\n\nOP2: from PC submit + ensemble - Version 6\nComplete · Antonina Dolgorukova · 2mo ago · mean(op2v8_0.614 + MLPv12_0.606 + pubNNv4_0.609)*0.45 + mean by drug * 0.45 + mean by ct * 0.1\n0.805\n\n0.598\n\nOP2: from PC submit + ensemble - Version 5\nError · Antonina Dolgorukova · 2mo ago · mean(op2v8_0.614 + MLPv12_0.606 + pubNNv4_0.609)*0.5 + mean by drug * 0.45 + mean by ct * 0.1\nOP2: from PC submit + ensemble - Version 4\nComplete · Antonina Dolgorukova · 2mo ago · mean(MLPv12_0.606 + pubNNv4_0.609)*0.45 + mean by drug * 0.45 + mean by ct * 0.1\n0.802\n\n0.597\n\nOP2: MLP submission - Version 28\nComplete · Antonina Dolgorukova · 2mo ago · with controls, train augm 0.1\n0.807\n\n0.613\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 124\nComplete · Alexander Chervov · 2mo ago · 3.0039 0.9682 LSVR onehot drug tsvd30 124 Opt2CV for each Target 3h15m\n0.853\n\n0.633\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 125\nComplete · Alexander Chervov · 2mo ago · 3.0106 0.9665 LSVR onehot drug tsvd30 125 OptAmbrosM for each target 2h51m\n0.853\n\n0.633\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 126\nComplete · Alexander Chervov · 2mo ago · 2.7093 1.0153 LSVR onehot drug tsvd30 126 OptMT for each target . 2h24m. C 1to1e4, eps mainly 1\n0.83\n\n0.631\n\nOP2: MLP submission - Version 25\nComplete · Antonina Dolgorukova · 2mo ago · control removed (?? mistake in the code); Drop 0.5*2\n0.85\n\n0.634\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 121\nComplete · Alexander Chervov · 2mo ago · 3.1446 0.9767 LSVR onehot drug tsvd30 121 Opt'AmbrosM' C=10,eps=2.5 one-prm-for-all starting from C=1,eps=1. 11min. 2rounds, but 1 is enough\n0.87\n\n0.643\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 120\nComplete · Alexander Chervov · 2mo ago · 2.7843 1.0323 LSVR onehot drug tsvd30 120 Opt'MT' C=1000,eps=10 one-prm-for-all starting from C=1,eps=1. 11min. 2rounds, but 1 is enough\n0.836\n\n0.635\n\nOP2: MLP submission - Version 24\nComplete · Antonina Dolgorukova · 2mo ago · control removed (?? mistake in the code), model as v12 + augm train + 0.1 noise\n0.805\n\n0.612\n\nOP2: MLP submission - Version 23\nComplete · Antonina Dolgorukova · 2mo ago · v12 tuned model, - controls, +train augm, + noise 0.1\n0.83\n\n0.622\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 89\nComplete · Alexander Chervov · 2mo ago · 2.6077 0.9943 50tsvd Ridge TE-ith , opt2CV try to better 0.611,0.612\n0.815\n\n0.61\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 91\nComplete · Alexander Chervov · 2mo ago · 0.9946 2.6085 tsvd35 Ridge te-ith opt2cv try to improve 0.612 by more components\n0.815\n\n0.611\n\nOP2: MLP submission - Version 20\nComplete · Antonina Dolgorukova · 2mo ago · v12, controls are excluded\n0.808\n\n0.626\n\nOP2: MLP submission - Version 19\nComplete · Antonina Dolgorukova · 2mo ago · as v12 but valid on rand, 5-fold\n0.801\n\n0.616\n\nOP2: MLP submission - Version 18\nComplete · Antonina Dolgorukova · 2mo ago · v12 + n_cells\n1.008\n\n0.73\n\nOP2: MLP submission - Version 16\nComplete · Antonina Dolgorukova · 2mo ago · dhid 256\n0.891\n\n0.658\n\nOP2: MLP submission - Version 15\nComplete · Antonina Dolgorukova · 2mo ago · 100/100 + n_cells, valid on test, 5 folds, dhid 64\n0.902\n\n0.665\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 75\nComplete · Alexander Chervov · 2mo ago · 0.9724 2.7624 lsvr te-i-th-tsvd25 opt2cv only C\n0.839\n\n0.621\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 83\nComplete · Alexander Chervov · 2mo ago · Lsvr te-ith-tsvd25 opt 4rnds C, Sm\n0.839\n\n0.619\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 86\nComplete · Alexander Chervov · 2mo ago · Try improve 0.612 te-ith-tsvd2 Ridge opt 2 cv\n0.817\n\n0.612\n\nOP2: MLP submission - Version 14\nComplete · Antonina Dolgorukova · 2mo ago · v 12, rrelu\n0.803\n\n0.608\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 74\nComplete · Alexander Chervov · 2mo ago · 0.975566 2.837578 target_enc_i_th_targetOpt_LVSRc1d100 tsvd25\n0.844\n\n0.62\n\nOP2: BL submission - Version 13\nComplete · Antonina Dolgorukova · 2mo ago · n_cells, 10 t PC\n0.844\n\n0.625\n\nOP2: BL submission - Version 12\nComplete · Antonina Dolgorukova · 2mo ago · n_cells 7 folds, 25 t PC\n0.843\n\n0.624\n\nOP2: MLP submission - Version 13\nComplete · Antonina Dolgorukova · 2mo ago · valid on NK\n0.827\n\n0.62\n\nOP2: BL submission - Version 10\nComplete · Antonina Dolgorukova · 2mo ago · n_cells, no folds, 25 t PC\n0.849\n\n0.63\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 70\nComplete · Alexander Chervov · 2mo ago · 1.009924 2.589438 target_enc_i_th_targetOpt_clipAlpha1e5\n0.828\n\n0.615\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 63\nComplete · Alexander Chervov · 2mo ago · 63 opt2CV 7h44 smooth0-4 [0.997175 2.573051] [0.9972 2.5731] 'smoothing4': [0,1e1,1e2] alpha = 2e4\n0.829\n\n0.62\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 69\nComplete · Alexander Chervov · 2mo ago · opt2CV [0.9953 2.5807] 8h40m alph, sm, 1round , 'alpha': [1e4,1e5,1e3], 'smoothing0': [1e1,1e2,0]\n0.823\n\n0.619\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 64\nComplete · Alexander Chervov · 2mo ago · LB0.638 64 opt2CV 9h34 alpha smooth0-4 [0.989163 2.476522] [0.9892 2.4765] 'smoothing4': [0,1e1,1e2], 'alpha': [1e3,1e4,1e5]}\n0.846\n\n0.638\n\nOP2: MLP submission - Version 12\nComplete · Antonina Dolgorukova · 2mo ago\n0.795\n\n0.606\n\nOP2: MLP submission - Version 11\nComplete · Antonina Dolgorukova · 2mo ago\n0.803\n\n0.616\n\nOP2: MLP submission - Version 9\nComplete · Antonina Dolgorukova · 2mo ago\n0.8\n\n0.611\n\nOP2: MLP submission - Version 9\nComplete · Antonina Dolgorukova · 2mo ago\n0.846\n\n0.641\n\nOP2: glmnet drug submission - Version 12\nComplete · Antonina Dolgorukova · 2mo ago · TE drug, pc t 145, pc f 50, 10 rand folds (10/90), thresh 0.35 rmse & < 0.1 overfit\n0.82\n\n0.614\n\nOP2: MLP submission - Version 8\nError · Antonina Dolgorukova · 2mo ago\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 56\nComplete · Alexander Chervov · 2mo ago · CVmt2.5817 CVamr1.0095 alpha optimized TE-all-tsvd25 drug , Ridge\n0.824\n\n0.621\n\n🧬OP2 Models,CV,Tuning - Version 81\nComplete · Alexander Chervov · 2mo ago · first blend 4top models 0.612-0.615 get LB0.610, and then blend with priors\n0.811\n\n0.604\n\n🧬OP2 Models,CV,Tuning - Version 80\nComplete · Alexander Chervov · 2mo ago · 0612-0615 blend simple submits precomputed\n0.816\n\n0.61\n\n🧬OP2 Models,CV,Tuning - Version 79\nComplete · Alexander Chervov · 2mo ago · blend 4 precompute LB 0604-605 submits\n0.812\n\n0.604\n\n🧬OP2 Models,CV,Tuning - Version 78\nComplete · Alexander Chervov · 2mo ago · Blend all 9 precomputed submits random+priors\n0.811\n\n0.604\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 39\nComplete · Alexander Chervov · 2mo ago · AmbrCV1.068 MTCV2.5491 Optimize MT TE for i-th target both features V39\n0.855\n\n0.633\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 41\nComplete · Alexander Chervov · 2mo ago · Amb1.000191 MT2.64004 TE for all drug targets same as previous other Alpha\n0.824\n\n0.616\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 45\nComplete · Alexander Chervov · 2mo ago · CV099415 TE for all tsvd30 optimized A,SM same for all\n0.834\n\n0.622\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 43\nComplete · Alexander Chervov · 2mo ago · CVA0.9931 CVM2.6365 reproduce(V75) 0.617 TE for ALL tsvd25 Ridge alpha, smooth chosen from prev optimization\n0.822\n\n0.617\n\nOP2: from PC submit + ensemble - Version 3\nComplete · Antonina Dolgorukova · 2mo ago · mean(opv8 0.614, pubNN v4 0.609) + drug, ct, 0.45/0.45/0.1\n0.81\n\n0.599\n\nOP2: Tuned public NN - Version 7\nComplete · Antonina Dolgorukova · 2mo ago · averaging 5 folds\n0.829\n\n0.612\n\nOP2: Tuned public NN - Version 5\nComplete · Antonina Dolgorukova · 2mo ago · early stop, patience 5\n0.837\n\n0.628\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 25\nComplete · Alexander Chervov · 2mo ago · CV0996 reproduce LB0612 Ridge on TE both drug and CT only i-th target tsvd25\n0.817\n\n0.612\n\nOP2: Tuned public NN - Version 4\nComplete · Antonina Dolgorukova · 2mo ago · custom_mean_rowwise_rmse loss, lr = 001\n0.828\n\n0.609\n\nOP2: Tuned public NN - Version 2\nComplete · Antonina Dolgorukova · 2mo ago · wrong data prep\n1.266\n\n0.897\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 23\nComplete · Alexander Chervov · 2mo ago · CV 0.2607 tsvd2 outstanding result in MT Ridge 0.1 onehot drug only\n0.83\n\n0.625\n\nOP2: glmnet drug submission - Version 11\nComplete · Antonina Dolgorukova · 2mo ago · TE drug, pc t 145, pc f 50, 10 rand folds, thresh 0.5 rmse & < 0.2 overfit\n0.818\n\n0.614\n\nOP2: glmnet drug submission - Version 10\nComplete · Antonina Dolgorukova · 2mo ago · TE drug, pc t 145, pc f 20, 10 rand folds, thresh 0.5 rmse & < 0.2 overfit\n0.822\n\n0.615\n\nOP2: glmnet drug submission - Version 9\nComplete · Antonina Dolgorukova · 2mo ago · TE drug, pc t 145, pc f 20, 10 rand folds, thresh 0.5 rmse\n0.823\n\n0.616\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 20\nComplete · Alexander Chervov · 2mo ago · blend with priors LB0.613 - CV2.66 vs2.670 original MT, fit_intercept=True(chg) Ridge=1(chg from0.1) tsvd30, onehot only drug\n0.814\n\n0.605\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 20\nComplete · Alexander Chervov · 2mo ago · CV2.66 vs2.670 original MT, fit_intercept=True(chg) Ridge=1(chg from0.1) tsvd30, onehot only drug\n0.823\n\n0.613\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 19\nComplete · Alexander Chervov · 2mo ago · CV2.668 vs2.670 original MT, Ridge0.1 tsvd30, onehot only drug\n0.821\n\n0.615\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 17\nComplete · Alexander Chervov · 2mo ago · blend with priors MT solution Ridge0.1 onehot drug only tsvd30\n0.809\n\n0.604\n\nOP2: glmnet drug submission - Version 8\nComplete · Antonina Dolgorukova · 2mo ago · targets from pval PCA, pc t 70, pc f 20, 10 rand folds, thresh 0.9 rmse\n0.894\n\n0.673\n\nOP2: glmnet drug submission - Version 6\nComplete · Antonina Dolgorukova · 2mo ago · as op2 v8 (0.614) with seed =1\n0.829\n\n0.615\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 12\nComplete · Alexander Chervov · 2mo ago · reproduce mt ridge onehot compound tsvd30 tandom20\n0.822\n\n0.615\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 12\nComplete · Alexander Chervov · 2mo ago · reproduce mt ridge onehot compound tsvd30\n0.822\n\n0.615\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 12\nComplete · Alexander Chervov · 2mo ago · same random20 no priors\n0.82\n\n0.619\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 12\nComplete · Alexander Chervov · 2mo ago · same as before blend with priors\n0.975\n\n0.697\n\nOP2 Advanced Modeling Tuning FeatEngineering Etc - Version 12\nComplete · Alexander Chervov · 2mo ago · cv25573 mt tsvd30 onrhot drug ridge naive optimized ridge per eacj target\n0.822\n\n0.621\n\nOP2: from PC submit + ensemble - Version 2\nComplete · Antonina Dolgorukova · 2mo ago · op2 v8 (0.614) + ensemble 0.45/0.45/0.1\n0.813\n\n0.604\n\nOP2: glmnet drug submission - Version 7\nComplete · Antonina Dolgorukova · 2mo ago · no cell type, 146, targets PCA (70 pc), 20 f pc, cv.glmnet, alpha = 1, 10 random folds thr < 0.5 rmse, train 0.67\n0.823\n\n0.616\n\nOP2: from PC submit + ensemble - Version 1\nComplete · Antonina Dolgorukova · 2mo ago · sep_cv.glmnet, s0.1_d (146, no cell type) pc f 20 all targets, alpha = 0, random 5 folds, rmse 0.74\n0.826\n\n0.615\n\nOP2: glmnet drug submission - Version 5\nComplete · Antonina Dolgorukova · 2mo ago · no cell type, 146, targets PCA (70 pc), 20 f pc, cv.glmnet, alpha = 1, 10 random folds thr < 0.4 rmse, train 0.8\n0.823\n\n0.616\n\nOP2: glmnet drug submission - Version 4\nComplete · Antonina Dolgorukova · 2mo ago · no cell type, 146, targets PCA (15 pc), 20 f pc, cv.glmnet, alpha = 1, 3 folds\n0.824\n\n0.617\n\nOP2: glmnet drug submission - Version 3\nComplete · Antonina Dolgorukova · 2mo ago · no cell type, 146, targets PCA (70 pc), 20 f pc, cv.glmnet, alpha = 1, 3 folds\n0.823\n\n0.616\n\nOP2 tune - Version 46\nComplete · Alexander Chervov · 2mo ago · CV09656 V46 tsvd100 LSVR optimized each target onehot Drugs submission_tsvd100LSVReachtargetOpt_CV09656.csv\n0.858\n\n0.633\n\nOP2: glmnet drug+cell submission - Version 24\nComplete · Antonina Dolgorukova · 2mo ago · no cell type,\n0.834\n\n0.624\n\nOP2: glmnet drug+cell submission - Version 23\nComplete · Antonina Dolgorukova · 2mo ago · as v22, alpha = 0\n0.91\n\n0.664\n\nOP2 tune - Version 44\nComplete · Alexander Chervov · 2mo ago · CV09664 tsvd50LSVReachtargetOpt_ onehot, only drugs, AmbrosM CV\n0.858\n\n0.633\n\nOP2 tune - Version 42\nComplete · Alexander Chervov · 2mo ago · CV0.9674 LSVR tsvd25 opt each target\n0.859\n\n0.634\n\nOP2: BL submission - Version 9\nComplete · Antonina Dolgorukova · 2mo ago · mol desc 20 pc (ignore cell type), 10 target pc, alpha = 1, 7 folds\n0.902\n\n0.675\n\nOP2: glmnet drug+cell submission - Version 22\nComplete · Antonina Dolgorukova · 2mo ago · targ. enc (s0.1), cv.glmnet, features pc 20/20, targets pc 70, alpha = 1, split = 0.8, 7 folds\n0.847\n\n0.626\n\nOP2 tune - Version 27\nComplete · Alexander Chervov · 2mo ago · two lb0617 ridges half sum , one from here - onehot, another from notebook2 - many TE features for each target\n0.823\n\n0.616\n\nOP2 tune - Version 26\nComplete · Alexander Chervov · 2mo ago · blend svr and ridge lb0.639 and 0.617\n0.842\n\n0.624\n\nOP2 tune - Version 25\nComplete · Alexander Chervov · 2mo ago · ridge onehot each target optimized CV09887. only drug features\n0.825\n\n0.617\n\nOP2 tune - Version 24\nComplete · Alexander Chervov · 2mo ago · linsvr default eps0.1 cv098996\n0.885\n\n0.654\n\nOP2 tune - Version 23\nComplete · Alexander Chervov · 2mo ago · linsvrC20Epsilon2 CV09765\n0.866\n\n0.639\n\nOP2: glmnet drug+cell submission - Version 21\nComplete · Antonina Dolgorukova · 2mo ago · rmse_valid < 1.6; rand folds targ. enc (s0.1), cv.glmnet, features pc 20/20, targets pc 10, alpha = 1, split = 0.8, 7 folds\n0.842\n\n0.627\n\nOP2: glmnet drug+cell submission - Version 19\nComplete · Antonina Dolgorukova · 2mo ago · targ. enc (s0.1), cv.glmnet, features pc 5/5, targets pc 10, alpha = 1, split = 0.8, 7 folds\n0.903\n\n0.659\n\nOP2: glmnet drug+cell submission - Version 18\nComplete · Antonina Dolgorukova · 2mo ago · targ. enc (s0.1), cv.glmnet, features pc 5/5, targets pc 5, alpha = 1, split = 0.8, 7 folds\n0.905\n\n0.664\n\nOP2: glmnet drug+cell submission - Version 17\nComplete · Antonina Dolgorukova · 2mo ago · BLv8 (0.613) no ensemble BLv8: targ. enc (s0.1), cv.glmnet, features pc 20/20, targets pc 10, alpha = 1, split = 0.8, 7 folds\n0.85\n\n0.629\n\nOP2: glmnet drug+cell submission - Version 16\nComplete · Antonina Dolgorukova · 2mo ago · targ. enc (s0.1), cv.glmnet, features pc 150/10, targets pc 150, alpha = 1, split = 0.8, 7 folds\n0.844\n\n0.625\n\nOP2: submission - Version 15\nComplete · Antonina Dolgorukova · 2mo ago · v14/2\n0.875\n\n0.637\n\nneural_network_regression - Version 2\nComplete · Dmitriy Ershov · 2mo ago\n0.836\n\n0.618\n\nOP2: submission - Version 14\nComplete · Antonina Dolgorukova · 2mo ago · targ. enc (drug+cell), sep glmnet by targ, features pc 150/10, targets pc 150, alpha = 1, split = 0.8, no folds\n1.007\n\n0.723\n\nneural_network_regression - Version 1\nComplete · Dmitriy Ershov · 2mo ago\n0.801\n\n0.606\n\nOP2: submission - Version 12\nComplete · Antonina Dolgorukova · 2mo ago · as v7 (0.616), alfa = 1 V7: target-encoded features, s0.1 (no cell type), no folds, 20 PC separate cv.glmnet for each target, alpha = 0\n0.829\n\n0.618\n\nOP2: BL submission - Version 8\nComplete · Antonina Dolgorukova · 2mo ago · alpha = 1, 7 folds 10 PC targets (enc drug+cell type) 20 PC features, cv.glmnet + ensemble\n0.828\n\n0.613\n\n🧬OP2 Models,CV,Tuning - Version 76\nComplete · Alexander Chervov · 2mo ago · Ensemble models v75 (Ridge many TE) with previous top and with cell type, compound means\n0.812\n\n0.605\n\n🧬OP2 Models,CV,Tuning - Version 75\nComplete · Alexander Chervov · 2mo ago · CV 0.993 - Ridge on many TE features tsvd feat. Only compound\n0.822\n\n0.617\n\nOP2: BL submission - Version 7\nComplete · Antonina Dolgorukova · 2mo ago · alpha = 0, 7 folds, targets (enc drug+cell type), 20 PC features, cv.glmnet + glmnet + ensemble\n0.893\n\n0.647\n\n🧬OP2 Models,CV,Tuning - Version 73\nComplete · Alexander Chervov · 2mo ago · Ensemble LB0.612(RidgeTEtsvd25) with aggregates compound, cell_type\n0.813\n\n0.605\n\n🧬OP2 Models,CV,Tuning - Version 72\nError · Alexander Chervov · 2mo ago · Ensemble of 0.612(RidgeTE) with aggregates by cell type and compound\n🧬OP2 Models,CV,Tuning - Version 70\nComplete · Alexander Chervov · 2mo ago · CV 1.036 LGB optimized params\n1.029\n\n0.729\n\nOP2: MLP submission - Version 7\nComplete · Antonina Dolgorukova · 2mo ago · target enc. by drug, PC20, 7 rand folds, d_hid1 = 1024 wd = 0.5 bs = 16 maxlr = 0.1\n0.824\n\n0.616\n\n🧬OP2 Models,CV,Tuning - Version 62\nComplete · Alexander Chervov · 2mo ago · catboost CV 1.106 (AmbrosM CV scheme)\n0.911\n\n0.664\n\n🧬OP2 Models,CV,Tuning - Version 54\nComplete · Alexander Chervov · 2mo ago · CV0.996639 - same as v53, but blending 50 random folds on submission, (model - new params - alpha for Ridge and smoothign for TE)\n0.817\n\n0.612\n\n🧬OP2 Models,CV,Tuning - Version 53\nComplete · Alexander Chervov · 2mo ago · CV0.996639 - use newly found params: alphas for Ridge (V47) and smoothings for Target Encoder (V51,52)\n0.817\n\n0.612\n\n🧬OP2 Models,CV,Tuning - Version 48\nComplete · Alexander Chervov · 2mo ago · CV0.997829 new best alpha for current smooth\n0.819\n\n0.615\n\n🧬OP2 Models,CV,Tuning - Version 45\nComplete · Alexander Chervov · 2mo ago · submit - blend 50folds, model from V43 CV0.999888 - opt alpha, smooth\n0.82\n\n0.614\n\n🧬OP2 Models,CV,Tuning - Version 43\nComplete · Alexander Chervov · 2mo ago · CV 0.999888 Optimal smoothers for each component, Ridge, TE\n0.819\n\n0.615\n\n🧬OP2 Models,CV,Tuning - Version 33\nComplete · Alexander Chervov · 2mo ago · best alp for each comp, submit = train on all , TE, Ridge, smooth1e7\n0.838\n\n0.625\n\n🧬OP2 Models,CV,Tuning - Version 32\nComplete · Alexander Chervov · 2mo ago · best alpha for each compoment, TE, Ridge, CV 1.023748\n0.838\n\n0.624\n\nOP2: MLP submission - Version 5\nComplete · Antonina Dolgorukova · 2mo ago · dummy + best MLP\n1.354\n\n1.003\n\nOP2: submission - Version 10\nComplete · Antonina Dolgorukova · 2mo ago · dummy + best ridge\n1.12\n\n0.845\n\nOP2: submission - Version 9\nError · Antonina Dolgorukova · 2mo ago · dummy, sep cv.glmnet seed = 1, no folds, 20PC\nOP2: BL submission - Version 5\nComplete · Antonina Dolgorukova · 2mo ago · v4 + ensemble\n0.823\n\n0.608\n\nOP2: BL submission - Version 4\nComplete · Antonina Dolgorukova · 2mo ago · averaging by drugs, CV3\n0.843\n\n0.621\n\nOP2: MLP submission - Version 2\nComplete · Antonina Dolgorukova · 2mo ago · d_hid1 = 1024 wd = 0.5 bs = 256 num_epochs = 100 maxlr = 0.01 n_folds = 3\n0.836\n\n0.621\n\nOP2: MLP submission - Version 1\nComplete · Antonina Dolgorukova · 2mo ago · d_hid1 = 1024 wd = 0.5 bs = 16 num_epochs = 100 maxlr = 0.01 n_folds = 3\n0.827\n\n0.616\n\nOP2: submission - Version 8\nComplete · Antonina Dolgorukova · 2mo ago · same as v5, seed = 42 for cv, 3 folds\n0.83\n\n0.614\n\nOP2: submission - Version 7\nComplete · Antonina Dolgorukova · 2mo ago · same as v5, seed = 1\n0.83\n\n0.616\n\n🧬OP2 Pytorch Embeddings for Beginners - Version 4\nComplete · Alexander Chervov · 2mo ago · Embeds with local r2 0.4+\n0.923\n\n0.673\n\nOP2: submission - Version 6\nComplete · Antonina Dolgorukova · 2mo ago · pca_target_encoded_features.qs, no cell type, averaged by drug, no folds, 170 PC separate cv.glmnet for each target\n0.888\n\n0.65\n\nOP2: submission - Version 5\nComplete · Antonina Dolgorukova · 2mo ago · target-encoded features, s0.1 (no cell type), no folds, 20 PC separate cv.glmnet for each target, alpha = 0, seed(42)\n0.83\n\n0.619\n\nOP2: submission - Version 4\nComplete · Antonina Dolgorukova · 2mo ago · pca_target_encoded_features.qs, no cell type, averaged by drug, no folds, 20 PC.\n0.841\n\n0.624\n\nOP2: submission - Version 2\nComplete · Antonina Dolgorukova · 2mo ago · pca_target_encoded_features.qs, no cell type, averaged by drug, 5 folds, 20 PC.\n0.843\n\n0.623\n\nOP2: submission - Version 1\nComplete · Antonina Dolgorukova · 2mo ago · pca_target_encoded_features.qs, with cell type (dummy), 3 folds, 10 PC.\n0.909\n\n0.653\n\nOP2: BL submission - Version 3\nComplete · Antonina Dolgorukova · 2mo ago · pca_target_encoded_features.qs, no cell type, 5 folds, 20 PC.\n0.843\n\n0.624\n\n🧬OP2 Pytorch Embeddings for Beginners - Version 1\nComplete · Alexander Chervov · 2mo ago · Embedding around 0.3 r2 on test\n0.838\n\n0.635\n\n🧬OP2 Models,CV,Tuning - Version 22\nComplete · Alexander Chervov · 2mo ago · 22 1.7104 alpha20 20tsvd-denoise 30folds\n0.933\n\n0.672\n\n🧬OP2 Models,CV,Tuning - Version 5\nComplete · Alexander Chervov · 2mo ago · 614 folds (leave one out)\n0.82\n\n0.616\n\n🧬OP2 Models - Version 4\nComplete · Alexander Chervov · 2mo ago · 50 folds with corrected inverse pca - from each fold - its own - not the last one as before\n0.82\n\n0.616\n\n🧬OP2 Models - Version 3\nComplete · Alexander Chervov · 2mo ago · 50 folds with r2>0.01 condition\n0.824\n\n0.617\n\n🧬OP2 Models - Version 2\nComplete · Alexander Chervov · 2mo ago · 50 folds\n0.823\n\n0.617\n\n🧬OP2 Models - Version 1\nComplete · Alexander Chervov · 2mo ago · 3 folds mean(aggr)\n0.828\n\n0.623\n\n🧬OP2 EDA,Baseline(s) - Version 37\nComplete · Alexander Chervov · 2mo ago · 37 approach 5 - aggr by compound + Ridge\n0.97\n\n0.707\n\n🧬OP2 EDA,Baseline(s) - Version 34\nComplete · Alexander Chervov · 2mo ago · V34 iterations=3, depth = 6\n0.96\n\n0.688\n\n🧬OP2 EDA,Baseline(s) - Version 33\nComplete · Alexander Chervov · 2mo ago · Catboost iterations 10, depth = 6\n1.005\n\n0.709\n\n🧬OP2 EDA,Baseline(s) - Version 32\nComplete · Alexander Chervov · 3mo ago · Catboost depth2\n1.345\n\n0.912\n\n🧬OP2 EDA,Baseline(s) - Version 30\nComplete · Alexander Chervov · 3mo ago · 30 Catboost first draft\n1.337\n\n0.905\n\n🧬OP2 EDA,Baseline(s) - Version 27\nComplete · Alexander Chervov · 3mo ago · Ridge nCT1 nCD25 Al10 TSVD35 - more relaxed compounds embeds and alpha\n0.914\n\n0.659\n\n🧬OP2 EDA,Baseline(s) - Version 26\nComplete · Alexander Chervov · 3mo ago · Ridge nCT1 nCD10 Al100 TSVD35\n0.916\n\n0.668\n\n🧬OP2 EDA,Baseline(s) - Version 25\nComplete · Alexander Chervov · 3mo ago · Ridge nCT1 nCD5 Al100 TSVD35\n0.927\n\n0.677\n\n🧬OP2 EDA,Baseline(s) - Version 24\nComplete · Alexander Chervov · 3mo ago · Ridge nCT3 nCD10 Al100 TSVD35\n0.978\n\n0.702\n\n🧬OP2 EDA,Baseline(s) - Version 23\nComplete · Alexander Chervov · 3mo ago · Ridge nCT10 nCD35 Al1 TSVD35\n1.062\n\n0.747\n\n🧬OP2 EDA,Baseline(s) - Version 18\nComplete · Alexander Chervov · 3mo ago · V18 TSVD-35 denoising quantile 0.54\n0.817\n\n0.623\n\n🧬OP2 EDA,Baseline(s) - Version 15\nComplete · Alexander Chervov · 3mo ago · TSVD ncomp35 quantile 0.54\n0.821\n\n0.626\n\n🧬OP2 EDA,Baseline(s) - Version 14\nComplete · Alexander Chervov · 3mo ago · ICA n_comp=35 quantile=.54\n0.818\n\n0.624\n\n🧬OP2 EDA,Baseline(s) - Version 13\nComplete · Alexander Chervov · 3mo ago · Pca25\n0.818\n\n0.626\n\n🧬OP2 EDA,Baseline(s) - Version 12\nComplete · Alexander Chervov · 3mo ago · pca100 denoising + groupby compound , quantile60\n0.82\n\n0.627\n\n🧬OP2 EDA,Baseline(s) - Version 11\nError · Alexander Chervov · 3mo ago · pca100 denoising + groupby compounds (quantile60)\n🧬OP2 EDA,Baseline(s) - Version 9\nComplete · Alexander Chervov · 3mo ago · groupby by compound and quantile 0.6\n0.835\n\n0.638\n\n🧬OP2 EDA,Baseline(s) - Version 6\nComplete · Alexander Chervov · 3mo ago · quantile 0.7\n0.908\n\n0.666\n\n🧬OP2 EDA,Baseline(s) - Version 5\nComplete · Alexander Chervov · 3mo ago · quantile 0.6\n0.891\n\n0.657\n\n🧬OP2 EDA,Baseline(s) - Version 4\nComplete · Alexander Chervov · 3mo ago · Medians\n0.891\n\n0.659\n\n🧬OP2 EDA,Baseline(s) - Version 1\nComplete · Alexander Chervov · 3mo ago · Means\n0.91\n\n0.664\n''' ","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2023-12-02T12:25:48.424785Z","iopub.execute_input":"2023-12-02T12:25:48.425254Z","iopub.status.idle":"2023-12-02T12:25:48.505057Z","shell.execute_reply.started":"2023-12-02T12:25:48.425218Z","shell.execute_reply":"2023-12-02T12:25:48.503526Z"},"jupyter":{"source_hidden":true},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"str_submits_MultimodalSingleCellIntegration = '''\nMmSCel 🧬Blend Submit - Version 4\nComplete · Alexander Chervov · 1y ago\n0.764497\n\n0.812236\n\nblender - Version 2\nComplete · geraseva · 1y ago · Notebook blender | Version 2\n0.76309\n\n0.811074\n\nblender - Version 1\nComplete · geraseva · 1y ago · Notebook blender | Version 1\n0.762149\n\n0.810542\n\nnotebookc7c2a15b67 - Version 21\nComplete · geraseva · 1y ago · Notebook notebookc7c2a15b67 | Version 21\n0.757265\n\n0.806795\n\nMmSCel 🧬Blend Submit - Version 3\nComplete · Alexander Chervov · 1y ago\n0.764423\n\n0.812095\n\nnotebookc7c2a15b67 - Version 20\nComplete · geraseva · 1y ago · lgb\n0.764855\n\n0.809707\n\nMmSCel 🧬Blend 4Models - Version 2\nComplete · Alexander Chervov · 1y ago\n0.763632\n\n0.811929\n\nFork of MmSCel 🧬CrossValidation schemes_sim_featu - Version 1\nComplete · Anton Kostin · 1y ago\n0.756427\n\n0.809665\n\nMmSCel 🧬 CV&Modeling Advanced - Version 50\nComplete · Alexander Chervov · 1y ago\n0.757794\n\n0.80887\n\nMmSCel 🧬 CV&Modeling Advanced - Version 51\nComplete · Alexander Chervov · 1y ago\n0.758213\n\n0.809011\n\nMmSCel 🧬 CV&Modeling Advanced - Version 51\nComplete · Alexander Chervov · 1y ago\n0.758213\n\n0.809011\n\nMmSCel 🧬 CV&Modeling corr features - Version 2\nComplete · geraseva · 1y ago · ridge\n0.752141\n\n0.805506\n\nnotebookc7c2a15b67 - Version 19\nComplete · geraseva · 1y ago · last lgb params, no blend\n0.76686\n\n0.812519\n\nnotebookc7c2a15b67 - Version 18\nComplete · geraseva · 1y ago · New params\n0.76686\n\n0.812519\n\nMmSCel 🧬 CV&Modeling Advanced - Version 44\nComplete · Alexander Chervov · 1y ago\n0.751601\n\n0.80547\n\nMmSCel 🧬 CV&Modeling Advanced - Version 43\nComplete · Alexander Chervov · 1y ago\n0.752651\n\n0.805291\n\n[🥈LB_T15| MSCI Multiome] CatBoostRegressor - Version 2\nComplete · Anna Sai · 1y ago\n0.763079\n\n0.809666\n\nsubmission_cite_seq_minus_0.00001_with_removed_features_5.csv\nComplete · Anton Kostin · 1y ago\n0.754893\n\n0.809007\n\nsubmission_cite_seq_minus_0.0001_with_removed_features.csv\nComplete · Anton Kostin · 1y ago\n0.755493\n\n0.80914\n\nsubmission_cite_seq_minus_0.00001_with_removed_features.csv\nComplete · Anton Kostin · 1y ago\n0.755536\n\n0.809149\n\nsubmission_cite_seq_3_with_removed_features.csv\nComplete · Anton Kostin · 1y ago\n0.755502\n\n0.809183\n\nsubmission_cite_seq_semantic_similarity_features_with_removed_features_0.csv\nComplete · Anton Kostin · 1y ago\n0.750474\n\n0.809145\n\nsubmission_cite_seq_semantic_similarity_features_with_removed_features.csv\nComplete · Anton Kostin · 1y ago\n0.752651\n\n0.805291\n\nnotebookc7c2a15b67 - Version 16\nComplete · geraseva · 1y ago · lgbm ensembled\n0.76686\n\n0.812519\n\nnotebookc7c2a15b67 - Version 15\nComplete · geraseva · 1y ago · lgbm ensembled\n0.76686\n\n0.812519\n\nnotebookc7c2a15b67 - Version 13\nComplete · geraseva · 1y ago · allinone\n0.76686\n\n0.812519\n\nnotebookc7c2a15b67 - Version 14\nComplete · geraseva · 1y ago · allinone\n0.76686\n\n0.812519\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 41\nComplete · Anton Kostin · 1y ago\n0.755486\n\n0.809133\n\nnotebookc7c2a15b67 - Version 12\nComplete · geraseva · 1y ago · cite corr and denoised imp ridge blended with best cv model + keras\n0.755486\n\n0.809133\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 40\nComplete · Anton Kostin · 1y ago\n0.755486\n\n0.809133\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 38\nComplete · Anton Kostin · 1y ago\n0.753157\n\n0.806343\n\nMmSCel 🧬CrossValidation schemes 2aa986 - Version 6\nComplete · Anna Sai · 1y ago\n0.752599\n\n0.805275\n\nMmSCel 🧬CrossValidation schemes 2aa986 - Version 3\nComplete · Anna Sai · 1y ago\n0.752702\n\n0.805436\n\nMmSCel 🧬CrossValidation schemes 2aa986 - Version 1\nComplete · Anna Sai · 1y ago\n0.635052\n\n0.731876\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 36\nComplete · Anton Kostin · 1y ago\n0.755483\n\n0.809133\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 31\nComplete · Anton Kostin · 1y ago\n0.755775\n\n0.808993\n\nnotebookc7c2a15b67 - Version 10\nComplete · geraseva · 1y ago · cite lgbm corr and denoised imp + multi keras\n0.764747\n\n0.810995\n\nnotebookc7c2a15b67 - Version 11\nComplete · geraseva · 1y ago · cite lgbm corr and magic and dca + multi keras\n0.764747\n\n0.810995\n\nnotebookc7c2a15b67 - Version 9\nComplete · geraseva · 1y ago · cite ridge corr and denoised imp + multi keras\n0.764747\n\n0.810995\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 28\nComplete · Anton Kostin · 1y ago\n0.755666\n\n0.809034\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 26\nComplete · Anton Kostin · 1y ago\n0.749627\n\n0.807023\n\nMmSCel 🧬CrossValidation schemes - Version 5\nComplete · Anton Kostin · 1y ago\n0.758286\n\n0.806801\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 24\nComplete · Anton Kostin · 1y ago\n0.75552\n\n0.809\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 22\nComplete · Anton Kostin · 1y ago\n0.75532\n\n0.808936\n\nnotebookc7c2a15b67 - Version 8\nComplete · geraseva · 1y ago · Notebook notebookc7c2a15b67 | multimodal keras + citeseq corr lgbm\n0.764747\n\n0.810995\n\nnotebookc7c2a15b67 - Version 7\nComplete · geraseva · 1y ago · Notebook notebookc7c2a15b67 | keras+ridge\n0.764747\n\n0.810995\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 20\nComplete · Anton Kostin · 1y ago\n0.75517\n\n0.808773\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 19\nComplete · Anton Kostin · 1y ago\n0.749056\n\n0.807273\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 18\nComplete · Anton Kostin · 1y ago\n0.73957\n\n0.807294\n\nMmSCel 🧬CrossValidation schemes - Version 3\nComplete · Anton Kostin · 1y ago\n0.761317\n\n0.808257\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 16\nComplete · Anton Kostin · 1y ago · Notebook with additional features (cd_pathways)\n0.749056\n\n0.807273\n\nMmSCel 🧬CrossValidation schemes - Version 4\nComplete · evgen8323 · 1y ago · Ridge - MLP (add)\n0.761191\n\n0.808687\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 14\nComplete · Anton Kostin · 1y ago · Added selected features from https://www.kaggle.com/datasets/alexandervc/feature-shop-for-multimodal-singlecell-competition and https://www.kaggle.com/datasets/grac2h5/top-100-closest-to-target-genes-with-features\n0.74878\n\n0.807194\n\nMmSCel 🧬CrossValidation schemes - Version 1\nComplete · evgen8323 · 1y ago · Ridge()\n0.752492\n\n0.8052\n\nMmSCel 🧬KNNRegressor - Version 1\nComplete · geraseva · 1y ago · Notebook MmSCel 🧬KNNRegressor | Version 1\n0.747837\n\n0.800833\n\nMmSCel 🧬CrossValidation schemes_sim_features - Version 10\nComplete · Anton Kostin · 1y ago · notebook with additional features based on semantic similarity of genes descpription\n0.754513\n\n0.806607\n\ndca_feature_selection - Version 2\nComplete · geraseva · 1y ago · Notebook dca_feature_selection | lgb\n0.731973\n\n0.792324\n\ndca_feature_selection - Version 1\nComplete · geraseva · 1y ago · Notebook dca_feature_selection | ridge\n0.73037\n\n0.782063\n\nMmSCel 🧬CrossValidation schemes - Version 27\nComplete · Alexander Chervov · 1y ago\n0.753024\n\n0.805882\n\nMmSCel 🧬CrossValidation schemes - Version 24\nComplete · Alexander Chervov · 1y ago\n0.752731\n\n0.805736\n\nMmSCel 🧬CrossValidation schemes - Version 22\nComplete · Alexander Chervov · 1y ago\n0.752646\n\n0.80565\n\nMmSCel 🧬CrossValidation schemes - Version 14\nComplete · Alexander Chervov · 1y ago\n0.752702\n\n0.805436\n\nMmSCel 🧬CrossValidation schemes - Version 13\nComplete · Alexander Chervov · 1y ago\n0.752536\n\n0.805292\n\nMmSCel 🧬CrossValidation schemes - Version 12\nComplete · Alexander Chervov · 1y ago\n0.75248\n\n0.805161\n\nMmSCel 🧬Experiments Modeling CrossValidation - Version 5\nComplete · Alexander Chervov · 1y ago\n0.746685\n\n0.805221\n\nMmSCel 🧬CrossValidation schemes - Version 11\nComplete · Alexander Chervov · 1y ago\n0.752487\n\n0.805139\n\nMmSCel 🧬CrossValidation schemes - Version 10\nComplete · Alexander Chervov · 1y ago\n0.752641\n\n0.80516\n\nMmSCel 🧬CrossValidation schemes - Version 9\nComplete · Alexander Chervov · 1y ago\n0.752589\n\n0.805137\n\nMmSCel 🧬CrossValidation schemes - Version 8\nComplete · Alexander Chervov · 1y ago\n0.75151\n\n0.803576\n\nMmSCel 🧬CrossValidation schemes - Version 7\nComplete · Alexander Chervov · 1y ago\n0.346729\n\n0.242511\n\nnotebookc7c2a15b67 - Version 4\nComplete · geraseva · 1y ago · cite: corr+dart; multi: linear on pytorch\n0.742583\n\n0.79839\n\nFork of magic_feature_selection - Version 8\nComplete · geraseva · 1y ago · magic dart knn 7\n0.723081\n\n0.790435\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 16\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 16\n0.286719\n\n0.215643\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 15\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 15\n-1\n\n0.17085\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 14\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 14\n-1\n\n0.218458\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 13\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 13\n0.230013\n\n0.168438\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 11\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 11\n0.230118\n\n0.168528\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 9\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 9\n0.230103\n\n0.168531\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 8\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 8\n0.22998\n\n0.16848\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 7\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 7\n0.229678\n\n0.168349\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 6\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 6\n0.229892\n\n0.168457\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 4\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 4\n0.228897\n\n0.167842\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 3\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 3\n0.216348\n\n0.158314\n\nMmSCel 🧬Analysis of EVALUATION and SUBMISSION - Version 2\nComplete · Alexander Chervov · 1y ago · Notebook MmSCel 🧬Analysis of EVALUATION and SUBMISSION | Version 2\n0.229995\n\n0.168397\n\nFork of magic_feature_selection - Version 6\nComplete · geraseva · 1y ago · denoising+PCA+lgb\n0.719044\n\n0.786285\n\nFork of magic_feature_selection - Version 5\nComplete · geraseva · 1y ago · Denoised + dart\n0.725089\n\n0.792015\n\nmagic_feature_selection - Version 24\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 24\n0.732563\n\n0.796325\n\nmagic_feature_selection - Version 23\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 23\n0.541254\n\n0.639518\n\nFork of magic_feature_selection - Version 4\nComplete · geraseva · 1y ago · denoised dataset + lgbm\n0.720732\n\n0.788543\n\nFork of magic_feature_selection - Version 3\nComplete · geraseva · 1y ago · denoised_dataset+ridge\n0.725222\n\n0.792011\n\nMSCI CITEseq Keras Quickstart - Version 1\nComplete · Alexander Chervov · 1y ago · Notebook MSCI CITEseq Keras Quickstart | Version 1\n0.76213\n\n0.809188\n\nmagic_feature_selection - Version 21\nComplete · geraseva · 1y ago · magic features + dart\n0.716049\n\n0.784666\n\nnotebookc7c2a15b67 - Version 3\nComplete · geraseva · 1y ago · corr features with dart\n0.732942\n\n0.796504\n\nmagic_feature_selection - Version 20\nComplete · geraseva · 1y ago · corr_features\n0.460811\n\n0.540954\n\nFork of magic_feature_selection - Version 1\nComplete · geraseva · 1y ago · full magic + feature selection + ridge\n0.718199\n\n0.784694\n\nFork of magic_feature_selection - Version 2\nComplete · geraseva · 1y ago · full magic 20 pc\n0.71002\n\n0.782182\n\nnotebookc7c2a15b67 - Version 2\nComplete · geraseva · 1y ago · Notebook notebookc7c2a15b67 | Version 2\n0.732621\n\n0.796578\n\nmagic_feature_selection - Version 18\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 18\n0.724375\n\n0.790192\n\nmagic_feature_selection - Version 17\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 17\n0.725385\n\n0.7927\n\nmagic_feature_selection - Version 16\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 16\n0.719324\n\n0.78686\n\nmagic_feature_selection - Version 15\nComplete · geraseva · 1y ago · Notebook magic_feature_selection | Version 15\n0.721579\n\n0.789611\n\nnotebook_for_feature_extraction - Version 13\nComplete · geraseva · 1y ago · Notebook notebook_for_feature_extraction | Version 13\n0.699706\n\n0.782094\n\nmagic+ridge - Version 14\nComplete · geraseva · 1y ago · magic+ridge\n0.708227\n\n0.7855\n\nciteseq_lgbm - Version 7\nComplete · geraseva · 1y ago · use minmaxscaler and ridge\n0.720968\n\n0.792514\n\nciteseq_lgbm - Version 6\nComplete · geraseva · 1y ago · Notebook citeseq_lgbm | Version 6\n0.728278\n\n0.794191\n\nciteseq_lgbm - Version 5\nComplete · geraseva · 1y ago · Notebook citeseq_lgbm | Version 5\n0.720979\n\n0.792439\n\nciteseq_lgbm - Version 4\nComplete · geraseva · 1y ago · Notebook citeseq_lgbm | Version 4\n0.719451\n\n0.791654\n\nopen_problem_multimodal - Version 2\nComplete · geraseva · 1y ago · Notebook open_problem_multimodal | Version 2\n0.709362\n\n0.784556\n\nMSCI CITEseq Quickstart - Version 5\nComplete · Alexander Chervov · 1y ago · Notebook MSCI CITEseq Quickstart | Version 5\n0.717348\n\n0.789131\n\nMSCI CITEseq Quickstart - Version 4\nComplete · Alexander Chervov · 1y ago · Notebook MSCI CITEseq Quickstart | Version 4\n0.717092\n\n0.789376\n\nMSCI CITEseq Quickstart - Version 1\nComplete · Alexander Chervov · 1y ago · Notebook MSCI CITEseq Quickstart | Version 1\n0.716826\n\n0.789392\n\nSimple Submission - Average by gene_id - Version 3\nComplete · Alexander Chervov · 1y ago · Notebook Simple Submission - Average by gene_id | Version 3\n0.615153\n\n0.719325\n\nSimple Submission - Average by gene_id - Version 2\nComplete · Alexander Chervov · 1y ago · Notebook Simple Submission - Average by gene_id | Version 2\n0.614936\n\n0.718893\n'''","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2023-12-02T12:25:48.507281Z","iopub.execute_input":"2023-12-02T12:25:48.508138Z","iopub.status.idle":"2023-12-02T12:25:48.533941Z","shell.execute_reply.started":"2023-12-02T12:25:48.508105Z","shell.execute_reply":"2023-12-02T12:25:48.532250Z"},"jupyter":{"source_hidden":true},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Parsing","metadata":{}},{"cell_type":"code","source":"import numpy as np\nimport pandas as pd\n\ns = str_raw_text\nif str_challenge_name == 'Multimodal Single-Cell Integration':\n    s = str_submits_MultimodalSingleCellIntegration\n\n\nl = s.split('\\n')[1:]\nl = [t for t in l if t != '']\nprint( l[:10] )\nres_private = []\nres_public =[]\nres_id1 = []\nres_id2 = []\nfor i,k in enumerate(l):# [:10]):\n    if k.startswith('Complete'):\n        #print(k)\n        res_id1.append(l[i-1])\n        res_id2.append(l[i])\n        res_private.append(float(l[i+1]))\n        res_public.append(float(l[i+2]))\n    \ndf = pd.DataFrame()\ndf['Public'] = res_public\ndf['Private'] = res_private\ndf['Notebook'] = res_id1\ndf['Id'] = res_id2\n\npd.set_option('display.max_columns', 500)\npd.set_option('display.max_rows', 500)\npd.set_option('display.width', 1000)\npd.set_option('display.max_colwidth', 1000)\n\n# Filtering\nm  = df[df.columns[0]] > 0; df = df[m]\nm  = df[df.columns[1]] > 0; df = df[m]\n\ndf.to_csv('df_stat_submissions.csv')\ndisplay(df.head(5)    )\n","metadata":{"execution":{"iopub.status.busy":"2023-12-02T12:25:48.535456Z","iopub.execute_input":"2023-12-02T12:25:48.535841Z","iopub.status.idle":"2023-12-02T12:25:48.573520Z","shell.execute_reply.started":"2023-12-02T12:25:48.535808Z","shell.execute_reply":"2023-12-02T12:25:48.572246Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Correlations","metadata":{}},{"cell_type":"code","source":"print('Pearson')\ndisplay( df.iloc[:,:2].round(2).corr() )\n\nprint('Spearman')\ndisplay( df.iloc[:,:2].round(2).corr(method = 'spearman') )\n\n\n# print('')\n# print('------------------ Without 2 bad publics ----------------- ')\n# print('')\n# d2 = df.sort_values(df.columns[0]  )\n# display(d2.head(4))\n# d2 = df.sort_values(df.columns[0]  ).iloc[2:,:]\n\n# print('Pearson')\n# display( d2.iloc[:,:2].round(2).corr() )\n\n# print('Spearman')\n# display( d2.iloc[:,:2].round(2).corr(method = 'spearman') )\n","metadata":{"execution":{"iopub.status.busy":"2023-12-02T12:25:48.575433Z","iopub.execute_input":"2023-12-02T12:25:48.575881Z","iopub.status.idle":"2023-12-02T12:25:48.598950Z","shell.execute_reply.started":"2023-12-02T12:25:48.575839Z","shell.execute_reply":"2023-12-02T12:25:48.597792Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Scatterplot","metadata":{}},{"cell_type":"code","source":"import seaborn as sns\nimport matplotlib.pyplot as plt\n\nplt.figure(figsize = (20,8))\n\nsns.scatterplot(x = df['Public'], y=df['Private'])\nplt.grid()\n#plt.legend(fontsize = 20 )\nplt.xlabel('Public', fontsize = 20)\nplt.ylabel('Private', fontsize = 20)\n\nc =  np.round( np.corrcoef(df.iloc[:,0],  df.iloc[:,1] )[0,1]  ,2)\nstr_corr = 'Correlation: ' + str(c)\nstr_title = 'Scores ' + str_challenge_name + ' challenge. Team: '+ str_team  + ' ' + str_corr\nplt.title(str_title, fontsize = 20)\nplt.show()\nprint('Two leftest points - are blends with \"bad\" (working only for public LB) public notebooks  ')\ndf.sort_values(df.columns[0]).head(10)","metadata":{"execution":{"iopub.status.busy":"2023-12-02T12:25:48.601413Z","iopub.execute_input":"2023-12-02T12:25:48.601767Z","iopub.status.idle":"2023-12-02T12:25:49.045759Z","shell.execute_reply.started":"2023-12-02T12:25:48.601738Z","shell.execute_reply":"2023-12-02T12:25:49.043986Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import matplotlib.pyplot as plt\n\n\nplt.plot(df['Public'], label = 'Public')\nplt.plot(df['Private'], label = 'Private')\nplt.grid()\nplt.legend( fontsize = 20 )\nc =  np.round( np.corrcoef(df.iloc[:,0],  df.iloc[:,1] )[0,1]  ,2)\nstr_corr = 'Correlation: ' + str(c)\nstr_title = 'Scores ' + str_challenge_name + ' challenge. Team: '+ str_team  + ' ' + str_corr\nplt.title(str_title, fontsize = 20)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2023-12-02T12:25:49.047730Z","iopub.execute_input":"2023-12-02T12:25:49.048061Z","iopub.status.idle":"2023-12-02T12:25:49.451653Z","shell.execute_reply.started":"2023-12-02T12:25:49.048032Z","shell.execute_reply":"2023-12-02T12:25:49.450534Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}