![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2F5c77063ce282211b9e0d6cb08e65f4a5%2Fphoto_2023-12-08_10-04-00.jpg?generation=1702029847880152&alt=media\")
š¹ Video: https://youtu.be/5xRxuDh_cGk
\nš Presentation: https://t.me/sberlogacompete/10211
Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
\nWould be happy for retwit
Gradient Boosted Decision Tree (GBDT) is a widely-used machine learning algorithm that has been shown to achieve state-of-the-art results on many standard data science problems. We are interested in its application to multioutput problems when the output is highly multidimensional. Although there are highly effective GBDT implementations, their scalability to such problems is still unsatisfactory. In this paper, we propose novel methods aiming to accelerate the training process of GBDT in the multioutput scenario. The idea behind these methods lies in the approximate computation of a scoring function used to find the best split of decision trees. These methods are implemented in SketchBoost, which itself is integrated into our easily customizable Python-based GPU implementation of GBDT called Py-Boost. Our numerical study demonstrates that SketchBoost speeds up the training process of GBDT by up to over 40 times while achieving comparable or even better performance.
\nIt easy to install: pip install py-boost
\nIt easy to use - see tutorial notebooks: Kaggle Open problems notebook, Tutorial_1_Basics, Tutorial_2_Advanced_multioutput, Tutorial_3_Custom_features
\n\nPaper: Iosipoi, Leonid, and Anton Vakhrushev. \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems.\" Advances in Neural Information Processing Systems 35 (2022): 25422-25435.
\nGold medals on Kaggle: CAFA5 (https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/434064) , Open problems - single cell perturbations 2023 (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/460191), Open problems 2022 (https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366471),
\nLots of silver/bronze medals in recent Open problems 2023 were based on Pyboost.
š¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Antonina will share outcomes of here analysis based on the following notebook:
\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat
On twitter: https://x.com/sberloga/status/1720895499559985164?s=20
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
š¹ Video: https://youtu.be/6ySKxnjHX8Y?si=ICiv26d7j1LbYaKn
\nšØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"
\nāļø Tuesday 17 October, 17.30 (CET time)
Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found link1, link2, link3.
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
==============================================================
\nš¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08
\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing
\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
Everybody is welcome to informal webinar around the competition. Everybody is welcome to share experience or take part in the discussion. Video records will be placed on youtube soon afterwards. Some webinars related to the last year \"Open problems\" Kaggle competition and several other challenges can be found here: https://www.youtube.com/playlist?list=PL1GnA8S_asBANHU95kOPOPEh3moPm5uhw
\nLet us start with some introductory webinar:
\nšØāš¬ Alexander Chervov \"Introduction to Kaggle competition - Single-Cell Perturbations (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations).\"
\nāļø Thursday 5 October, 18.00 (CET time, e.g. Paris )
We will give brief introduction mainly from machine learning perspective. Task contains 614 samples in train and 255 in test, with only two features, which both are categorical (cell type, and drug). Metric is MMSE (row wise). Cross-validation - by cell types with modifications. Baselines with encoding categorical features - one-hot, target encoding with optimization of smoothing parameter, pytorch embedding neural network. Alternative features: \"ChemBert\" ( link (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/441550) - by Aleksey Trepetsky), molecular descriptors ( link (https://www.kaggle.com/code/antoninadolgorukova/op2-feature-engineering) by Antonina Dolgorukova), etc. Most of approaches use dimensional reduction (pca-like) of targets (18211) to smaller dimension 25-100, then predicting these reduced target and then making inverse transform. Notebooks: EDA (https://www.kaggle.com/code/alexandervc/op2-eda-baseline-s) , modeling (https://www.kaggle.com/code/alexandervc/op2-models-cv-tuning) , pytorch embeddings (https://www.kaggle.com/code/alexandervc/op2-pytorch-embeddings-for-beginners) , single cell RNA-seq data brief look (https://www.kaggle.com/code/alexandervc/op2-rna-seq-data-scanpy-adata-cell-cycle).
\nUnusual: 50 000$$ - for biologically insightful solutions which will shed light on \"How did you integrate the ATAC data? Did you learn a gene regulatory network?\" etc.
Zoom link will be available here 5 minutes before the start.
", "messages": [ { "id": "2466416", "postDate": "10/03/2023 19:18:12", "content": "š¹ Video: https://youtu.be/5xRxuDh_cGk
\nš Presentation: https://t.me/sberlogacompete/10211
Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
\nWould be happy for retwit
Gradient Boosted Decision Tree (GBDT) is a widely-used machine learning algorithm that has been shown to achieve state-of-the-art results on many standard data science problems. We are interested in its application to multioutput problems when the output is highly multidimensional. Although there are highly effective GBDT implementations, their scalability to such problems is still unsatisfactory. In this paper, we propose novel methods aiming to accelerate the training process of GBDT in the multioutput scenario. The idea behind these methods lies in the approximate computation of a scoring function used to find the best split of decision trees. These methods are implemented in SketchBoost, which itself is integrated into our easily customizable Python-based GPU implementation of GBDT called Py-Boost. Our numerical study demonstrates that SketchBoost speeds up the training process of GBDT by up to over 40 times while achieving comparable or even better performance.
\nIt easy to install: pip install py-boost
\nIt easy to use - see tutorial notebooks: Kaggle Open problems notebook, Tutorial_1_Basics, Tutorial_2_Advanced_multioutput, Tutorial_3_Custom_features
\n\nPaper: Iosipoi, Leonid, and Anton Vakhrushev. \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems.\" Advances in Neural Information Processing Systems 35 (2022): 25422-25435.
\nGold medals on Kaggle: CAFA5 (https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/434064) , Open problems - single cell perturbations 2023 (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/460191), Open problems 2022 (https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366471),
\nLots of silver/bronze medals in recent Open problems 2023 were based on Pyboost.
š¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Antonina will share outcomes of here analysis based on the following notebook:
\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat
On twitter: https://x.com/sberloga/status/1720895499559985164?s=20
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
š¹ Video: https://youtu.be/6ySKxnjHX8Y?si=ICiv26d7j1LbYaKn
\nšØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"
\nāļø Tuesday 17 October, 17.30 (CET time)
Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found link1, link2, link3.
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
==============================================================
\nš¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08
\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing
\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
Everybody is welcome to informal webinar around the competition. Everybody is welcome to share experience or take part in the discussion. Video records will be placed on youtube soon afterwards. Some webinars related to the last year \"Open problems\" Kaggle competition and several other challenges can be found here: https://www.youtube.com/playlist?list=PL1GnA8S_asBANHU95kOPOPEh3moPm5uhw
\nLet us start with some introductory webinar:
\nšØāš¬ Alexander Chervov \"Introduction to Kaggle competition - Single-Cell Perturbations (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations).\"
\nāļø Thursday 5 October, 18.00 (CET time, e.g. Paris )
We will give brief introduction mainly from machine learning perspective. Task contains 614 samples in train and 255 in test, with only two features, which both are categorical (cell type, and drug). Metric is MMSE (row wise). Cross-validation - by cell types with modifications. Baselines with encoding categorical features - one-hot, target encoding with optimization of smoothing parameter, pytorch embedding neural network. Alternative features: \"ChemBert\" ( link (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/441550) - by Aleksey Trepetsky), molecular descriptors ( link (https://www.kaggle.com/code/antoninadolgorukova/op2-feature-engineering) by Antonina Dolgorukova), etc. Most of approaches use dimensional reduction (pca-like) of targets (18211) to smaller dimension 25-100, then predicting these reduced target and then making inverse transform. Notebooks: EDA (https://www.kaggle.com/code/alexandervc/op2-eda-baseline-s) , modeling (https://www.kaggle.com/code/alexandervc/op2-models-cv-tuning) , pytorch embeddings (https://www.kaggle.com/code/alexandervc/op2-pytorch-embeddings-for-beginners) , single cell RNA-seq data brief look (https://www.kaggle.com/code/alexandervc/op2-rna-seq-data-scanpy-adata-cell-cycle).
\nUnusual: 50 000$$ - for biologically insightful solutions which will shed light on \"How did you integrate the ATAC data? Did you learn a gene regulatory network?\" etc.
Zoom link will be available here 5 minutes before the start.
", "rawMarkdown": "š¹ Video: https://youtu.be/5xRxuDh_cGk\nš Presentation: https://t.me/sberlogacompete/10211\n\nPyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it: \n\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"\nāļø Monday 11 December 17.00 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231211T160000Z%2F20231211T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=Anton%20Vakhrushev%20%22SketchBoost%3A%20Fast%20Gradient%20Boosted%20Decision%20Tree%20for%20Multioutput%20Problems%22)\n\nTwitter: https://x.com/sberloga/status/1733589792024080479?s=20\nWould be happy for retwit \n\nGradient Boosted Decision Tree (GBDT) is a widely-used machine learning algorithm that has been shown to achieve state-of-the-art results on many standard data science problems. We are interested in its application to multioutput problems when the output is highly multidimensional. Although there are highly effective GBDT implementations, their scalability to such problems is still unsatisfactory. In this paper, we propose novel methods aiming to accelerate the training process of GBDT in the multioutput scenario. The idea behind these methods lies in the approximate computation of a scoring function used to find the best split of decision trees. These methods are implemented in SketchBoost, which itself is integrated into our easily customizable Python-based GPU implementation of GBDT called Py-Boost. Our numerical study demonstrates that SketchBoost speeds up the training process of GBDT by up to over 40 times while achieving comparable or even better performance.\n\nIt easy to install: [pip install py-boost](https://pypi.org/project/py-boost/)\n\nIt easy to use - see tutorial notebooks: [Kaggle Open problems notebook](https://www.kaggle.com/code/alexandervc/pyboost-secret-grandmaster-s-tool), [Tutorial_1_Basics](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_1_Basics.ipynb), [Tutorial_2_Advanced_multioutput](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_2_Advanced_multioutput.ipynb), [Tutorial_3_Custom_features](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_3_Custom_features.ipynb)\n\n[Github](https://github.com/sb-ai-lab/Py-Boost) \n\n[Paper](https://openreview.net/forum?id=WSxarC8t-T): Iosipoi, Leonid, and Anton Vakhrushev. \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems.\" Advances in Neural Information Processing Systems 35 (2022): 25422-25435. \n\nGold medals on Kaggle: CAFA5 (https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/434064) , Open problems - single cell perturbations 2023 (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/460191), Open problems 2022 (https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366471),\nLots of silver/bronze medals in recent Open problems 2023 were based on Pyboost. \n\n\n\n\n------------------------------------------------------------------------------------------------------------------------\n\nš¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka\n\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"\nāļø Monday 06 November, 17.00 (CET time)\n\n[Add to google calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231106T160000Z%2F20231106T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20Antonina%20Dolgorukova%20%22Single%20Cell%20Perturbations%20data%20analysis%22)\n\nAntonina will share outcomes of here analysis based on the following notebook:\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat\n\n- Calculation of Mitochondrial/ribosomal contamination (all cells), what percentage of all genes copies comes from the single most observed gene in each cell (10% of cells)\n- Identification of highly variable features, followed by PCA, selection of principles components (10% of cells)\n- Clustering cells based on main PC for the highly variable features\n- Running non-linear dimensional reduction (UMAP/tSNE)\n\n\nOn twitter: https://x.com/sberloga/status/1720895499559985164?s=20\n\nZoom link will be available here 5 minutes before the start.\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga\n\n--------------------------------------------------------------------\n\nš¹ Video: https://youtu.be/6ySKxnjHX8Y?si=ICiv26d7j1LbYaKn\n\nšØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"\nāļø Tuesday 17 October, 17.30 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231017T153000Z%2F20231017T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20%20%20Brainstorm%20on%20%20%22Kaggle%3A%20Open%20Problems%20-%20Single-Cell%20Perturbations.%22)\n\n Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found [link1](https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing), [link2] (https://docs.google.com/spreadsheets/d/1APN63PMaWZygVjYimK9Ivt0RvifdAU5JRYkxiDn4szw/edit?usp=sharing), [link3](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/445883).\n\nZoom link will be available here 5 minutes before the start. \nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga \n\n==============================================================\n\nš¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true \n\nEverybody is welcome to informal webinar around the competition. Everybody is welcome to share experience or take part in the discussion. Video records will be placed on youtube soon afterwards. Some webinars related to the last year \"Open problems\" Kaggle competition and several other challenges can be found here: https://www.youtube.com/playlist?list=PL1GnA8S_asBANHU95kOPOPEh3moPm5uhw \n\nLet us start with some introductory webinar: \n\nšØāš¬ Alexander Chervov \"Introduction to Kaggle competition - Single-Cell Perturbations (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations).\"\nāļø Thursday 5 October, 18.00 (CET time, e.g. Paris )\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231005T160000Z%2F20231005T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20Alexander%20Chervov%20%22Introduction%20to%20Kaggle%20competition%20-%20Single-Cell%20Perturbations)\n\nWe will give brief introduction mainly from machine learning perspective. Task contains 614 samples in train and 255 in test, with only two features, which both are categorical (cell type, and drug). Metric is MMSE (row wise). Cross-validation - by cell types with modifications. Baselines with encoding categorical features - one-hot, target encoding with optimization of smoothing parameter, pytorch embedding neural network. Alternative features: \"ChemBert\" ( link (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/441550) - by Aleksey Trepetsky), molecular descriptors ( link (https://www.kaggle.com/code/antoninadolgorukova/op2-feature-engineering) by Antonina Dolgorukova), etc. Most of approaches use dimensional reduction (pca-like) of targets (18211) to smaller dimension 25-100, then predicting these reduced target and then making inverse transform. Notebooks: EDA (https://www.kaggle.com/code/alexandervc/op2-eda-baseline-s) , modeling (https://www.kaggle.com/code/alexandervc/op2-models-cv-tuning) , pytorch embeddings (https://www.kaggle.com/code/alexandervc/op2-pytorch-embeddings-for-beginners) , single cell RNA-seq data brief look (https://www.kaggle.com/code/alexandervc/op2-rna-seq-data-scanpy-adata-cell-cycle).\nUnusual: 50 000$$ - for biologically insightful solutions which will shed light on \"How did you integrate the ATAC data? Did you learn a gene regulatory network?\" etc. \n\n\nZoom link will be available here 5 minutes before the start.", "votes": null }, { "id": "2467007", "postDate": "10/04/2023 08:32:29", "content": "Cool, thank you for organizing this!
", "rawMarkdown": "Cool, thank you for organizing this!", "votes": null }, { "id": "2468568", "postDate": "10/05/2023 15:57:24", "content": "Zoom:
\nhttps://us02web.zoom.us/j/87351847201?pwd=WmJwSHRET0M1c2ZRMHZIRHFpMHFWZz09
The presentation with molecular descriptors explained is here: https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
", "rawMarkdown": "The presentation with molecular descriptors explained is here: https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true", "votes": null }, { "id": "2469258", "postDate": "10/06/2023 08:40:37", "content": "š¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08
\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing
\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
For now, I have checked the descriptors as is + PCA, in this version of my baseline notebook - LB score is much worse than 0s (0.675).
\nIt confirms what I said, feature selection is needed here, followed by some standartization/normalization.
", "rawMarkdown": "For now, I have checked the descriptors as is + PCA, in [this version](https://www.kaggle.com/code/antoninadolgorukova/op2-bl-submission?scriptVersionId=145674979) of my baseline notebook - LB score is much worse than 0s (0.675).\n\nIt confirms what I said, feature selection is needed here, followed by some standartization/normalization.", "votes": null }, { "id": "2474411", "postDate": "10/09/2023 07:30:41", "content": "Thank you for the introductory webinar
", "rawMarkdown": "Thank you for the introductory webinar", "votes": null }, { "id": "2480625", "postDate": "10/13/2023 12:52:43", "content": "Thank you, very much! Alexander!
", "rawMarkdown": "Thank you, very much! Alexander!", "votes": null }, { "id": "2483559", "postDate": "10/15/2023 19:31:30", "content": "šØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"
\nāļø Tuesday 17 October, 17.30 (CET time)
Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found link1, link2, link3.
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
Zoom link:
\nhttps://us02web.zoom.us/j/81679339241?pwd=M0E2OWk0ZmJ6S2xLNGZoWlVJNEYvdz09
@alexandervc Are these meetings still happening?
", "rawMarkdown": "alexandervc Are these meetings still happening?", "votes": null }, { "id": "2489538", "postDate": "10/20/2023 04:13:36", "content": "There were two , hopefully there will be more , announces will be here
", "rawMarkdown": "There were two , hopefully there will be more , announces will be here", "votes": null }, { "id": "2512754", "postDate": "11/04/2023 19:51:25", "content": "šØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Antonina will share outcomes of here analysis based on the following notebook:
\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat
Zoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
Zoom link: https://us02web.zoom.us/j/89080226552?pwd=aGc1NElmWlVuZEljYjBaS1pZaUN2QT09
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
š¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Thanks for the webinars @alexandervc - sounds like a quick intro to the competition for people who are late enough :))
", "rawMarkdown": "Thanks for the webinars @alexandervc - sounds like a quick intro to the competition for people who are late enough :))", "votes": null }, { "id": "2553491", "postDate": "12/08/2023 09:48:59", "content": "Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
\nWould be happy for retwit
Zoom link:
\nhttps://us02web.zoom.us/j/86106703038?pwd=VFBtUm9HTW5wWmFUUTJnUG1scytOQT09
š¹ Video: https://youtu.be/5xRxuDh_cGk
\nš Presentation: https://t.me/sberlogacompete/10211
Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev, Leonid Iosipoi \"Pyboost/SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
", "rawMarkdown": "š¹ Video: https://youtu.be/5xRxuDh_cGk\nš Presentation: https://t.me/sberlogacompete/10211\n\nPyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:\n\nšØāš¬ Anton Vakhrushev, Leonid Iosipoi \"Pyboost/SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"\nāļø Monday 11 December 17.00 (CET time)\n\nTwitter: https://x.com/sberloga/status/1733589792024080479?s=20", "votes": null }, { "id": "2994862", "postDate": "09/21/2024 15:13:29", "content": "nice work!
", "rawMarkdown": "nice work!", "votes": null } ], "comments": [ { "id": 2467007, "author_name": "antoninadolgorukova", "author_url": "", "post_date": "10/04/2023 08:32:29", "content": "Cool, thank you for organizing this!
", "votes": null, "replies": [] }, { "id": 2468568, "author_name": "alexandervc", "author_url": "", "post_date": "10/05/2023 15:57:24", "content": "Zoom:
\nhttps://us02web.zoom.us/j/87351847201?pwd=WmJwSHRET0M1c2ZRMHZIRHFpMHFWZz09
Thank you, very much! Alexander!
", "votes": null, "replies": [] } ] }, { "id": 2468991, "author_name": "antoninadolgorukova", "author_url": "", "post_date": "10/06/2023 03:51:59", "content": "The presentation with molecular descriptors explained is here: https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
", "votes": null, "replies": [ { "id": 2474195, "author_name": "antoninadolgorukova", "author_url": "", "post_date": "10/09/2023 04:17:58", "content": "For now, I have checked the descriptors as is + PCA, in this version of my baseline notebook - LB score is much worse than 0s (0.675).
\nIt confirms what I said, feature selection is needed here, followed by some standartization/normalization.
", "votes": null, "replies": [] } ] }, { "id": 2469258, "author_name": "alexandervc", "author_url": "", "post_date": "10/06/2023 08:40:37", "content": "š¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08
\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing
\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true
Thank you for the introductory webinar
", "votes": null, "replies": [] }, { "id": 2483559, "author_name": "alexandervc", "author_url": "", "post_date": "10/15/2023 19:31:30", "content": "šØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"
\nāļø Tuesday 17 October, 17.30 (CET time)
Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found link1, link2, link3.
\nZoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
Zoom link:
\nhttps://us02web.zoom.us/j/81679339241?pwd=M0E2OWk0ZmJ6S2xLNGZoWlVJNEYvdz09
@alexandervc Are these meetings still happening?
", "votes": null, "replies": [ { "id": 2489538, "author_name": "alexandervc", "author_url": "", "post_date": "10/20/2023 04:13:36", "content": "There were two , hopefully there will be more , announces will be here
", "votes": null, "replies": [] } ] } ] }, { "id": 2512754, "author_name": "alexandervc", "author_url": "", "post_date": "11/04/2023 19:51:25", "content": "šØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Antonina will share outcomes of here analysis based on the following notebook:
\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat
Zoom link will be available here 5 minutes before the start.
\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga
Zoom link: https://us02web.zoom.us/j/89080226552?pwd=aGc1NElmWlVuZEljYjBaS1pZaUN2QT09
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
š¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka
\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"
\nāļø Monday 06 November, 17.00 (CET time)
Thanks for the webinars @alexandervc - sounds like a quick intro to the competition for people who are late enough :))
", "votes": null, "replies": [] }, { "id": 2553491, "author_name": "alexandervc", "author_url": "", "post_date": "12/08/2023 09:48:59", "content": "Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
\nWould be happy for retwit
Zoom link:
\nhttps://us02web.zoom.us/j/86106703038?pwd=VFBtUm9HTW5wWmFUUTJnUG1scytOQT09
š¹ Video: https://youtu.be/5xRxuDh_cGk
\nš Presentation: https://t.me/sberlogacompete/10211
Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:
\nšØāš¬ Anton Vakhrushev, Leonid Iosipoi \"Pyboost/SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"
\nāļø Monday 11 December 17.00 (CET time)
Twitter: https://x.com/sberloga/status/1733589792024080479?s=20
", "votes": null, "replies": [] }, { "id": 2994862, "author_name": "", "author_url": "", "post_date": "09/21/2024 15:13:29", "content": "nice work!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2466416": "š¹ Video: https://youtu.be/5xRxuDh_cGk\nš Presentation: https://t.me/sberlogacompete/10211\n\nPyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it: \n\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"\nāļø Monday 11 December 17.00 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231211T160000Z%2F20231211T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=Anton%20Vakhrushev%20%22SketchBoost%3A%20Fast%20Gradient%20Boosted%20Decision%20Tree%20for%20Multioutput%20Problems%22)\n\nTwitter: https://x.com/sberloga/status/1733589792024080479?s=20\nWould be happy for retwit \n\nGradient Boosted Decision Tree (GBDT) is a widely-used machine learning algorithm that has been shown to achieve state-of-the-art results on many standard data science problems. We are interested in its application to multioutput problems when the output is highly multidimensional. Although there are highly effective GBDT implementations, their scalability to such problems is still unsatisfactory. In this paper, we propose novel methods aiming to accelerate the training process of GBDT in the multioutput scenario. The idea behind these methods lies in the approximate computation of a scoring function used to find the best split of decision trees. These methods are implemented in SketchBoost, which itself is integrated into our easily customizable Python-based GPU implementation of GBDT called Py-Boost. Our numerical study demonstrates that SketchBoost speeds up the training process of GBDT by up to over 40 times while achieving comparable or even better performance.\n\nIt easy to install: [pip install py-boost](https://pypi.org/project/py-boost/)\n\nIt easy to use - see tutorial notebooks: [Kaggle Open problems notebook](https://www.kaggle.com/code/alexandervc/pyboost-secret-grandmaster-s-tool), [Tutorial_1_Basics](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_1_Basics.ipynb), [Tutorial_2_Advanced_multioutput](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_2_Advanced_multioutput.ipynb), [Tutorial_3_Custom_features](https://github.com/AILab-MLTools/Py-Boost/blob/master/tutorials/Tutorial_3_Custom_features.ipynb)\n\n[Github](https://github.com/sb-ai-lab/Py-Boost) \n\n[Paper](https://openreview.net/forum?id=WSxarC8t-T): Iosipoi, Leonid, and Anton Vakhrushev. \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems.\" Advances in Neural Information Processing Systems 35 (2022): 25422-25435. \n\nGold medals on Kaggle: CAFA5 (https://www.kaggle.com/competitions/cafa-5-protein-function-prediction/discussion/434064) , Open problems - single cell perturbations 2023 (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/460191), Open problems 2022 (https://www.kaggle.com/competitions/open-problems-multimodal/discussion/366471),\nLots of silver/bronze medals in recent Open problems 2023 were based on Pyboost. \n\n\n\n\n------------------------------------------------------------------------------------------------------------------------\n\nš¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka\n\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"\nāļø Monday 06 November, 17.00 (CET time)\n\n[Add to google calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231106T160000Z%2F20231106T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20Antonina%20Dolgorukova%20%22Single%20Cell%20Perturbations%20data%20analysis%22)\n\nAntonina will share outcomes of here analysis based on the following notebook:\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat\n\n- Calculation of Mitochondrial/ribosomal contamination (all cells), what percentage of all genes copies comes from the single most observed gene in each cell (10% of cells)\n- Identification of highly variable features, followed by PCA, selection of principles components (10% of cells)\n- Clustering cells based on main PC for the highly variable features\n- Running non-linear dimensional reduction (UMAP/tSNE)\n\n\nOn twitter: https://x.com/sberloga/status/1720895499559985164?s=20\n\nZoom link will be available here 5 minutes before the start.\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga\n\n--------------------------------------------------------------------\n\nš¹ Video: https://youtu.be/6ySKxnjHX8Y?si=ICiv26d7j1LbYaKn\n\nšØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"\nāļø Tuesday 17 October, 17.30 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231017T153000Z%2F20231017T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20%20%20Brainstorm%20on%20%20%22Kaggle%3A%20Open%20Problems%20-%20Single-Cell%20Perturbations.%22)\n\n Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found [link1](https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing), [link2] (https://docs.google.com/spreadsheets/d/1APN63PMaWZygVjYimK9Ivt0RvifdAU5JRYkxiDn4szw/edit?usp=sharing), [link3](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/445883).\n\nZoom link will be available here 5 minutes before the start. \nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga \n\n==============================================================\n\nš¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true \n\nEverybody is welcome to informal webinar around the competition. Everybody is welcome to share experience or take part in the discussion. Video records will be placed on youtube soon afterwards. Some webinars related to the last year \"Open problems\" Kaggle competition and several other challenges can be found here: https://www.youtube.com/playlist?list=PL1GnA8S_asBANHU95kOPOPEh3moPm5uhw \n\nLet us start with some introductory webinar: \n\nšØāš¬ Alexander Chervov \"Introduction to Kaggle competition - Single-Cell Perturbations (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations).\"\nāļø Thursday 5 October, 18.00 (CET time, e.g. Paris )\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231005T160000Z%2F20231005T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20Alexander%20Chervov%20%22Introduction%20to%20Kaggle%20competition%20-%20Single-Cell%20Perturbations)\n\nWe will give brief introduction mainly from machine learning perspective. Task contains 614 samples in train and 255 in test, with only two features, which both are categorical (cell type, and drug). Metric is MMSE (row wise). Cross-validation - by cell types with modifications. Baselines with encoding categorical features - one-hot, target encoding with optimization of smoothing parameter, pytorch embedding neural network. Alternative features: \"ChemBert\" ( link (https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/441550) - by Aleksey Trepetsky), molecular descriptors ( link (https://www.kaggle.com/code/antoninadolgorukova/op2-feature-engineering) by Antonina Dolgorukova), etc. Most of approaches use dimensional reduction (pca-like) of targets (18211) to smaller dimension 25-100, then predicting these reduced target and then making inverse transform. Notebooks: EDA (https://www.kaggle.com/code/alexandervc/op2-eda-baseline-s) , modeling (https://www.kaggle.com/code/alexandervc/op2-models-cv-tuning) , pytorch embeddings (https://www.kaggle.com/code/alexandervc/op2-pytorch-embeddings-for-beginners) , single cell RNA-seq data brief look (https://www.kaggle.com/code/alexandervc/op2-rna-seq-data-scanpy-adata-cell-cycle).\nUnusual: 50 000$$ - for biologically insightful solutions which will shed light on \"How did you integrate the ATAC data? Did you learn a gene regulatory network?\" etc. \n\n\nZoom link will be available here 5 minutes before the start.", "2467007": "Cool, thank you for organizing this!", "2468568": "Zoom:\nhttps://us02web.zoom.us/j/87351847201?pwd=WmJwSHRET0M1c2ZRMHZIRHFpMHFWZz09", "2468991": "The presentation with molecular descriptors explained is here: https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true", "2469258": "š¹ Video: https://youtu.be/dRG3qTaALp0?si=c6k6CARJuik-xF08\nš Presentation: A.Chervov: https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing\nš Presentation: A.Dolgorukova \"Molecular descriptors\": https://docs.google.com/presentation/d/1H7qL73UrRY3tjpSM3aPxIgETBeMkKzid/edit?usp=sharing&ouid=114126852933828504381&rtpof=true&sd=true", "2474195": "For now, I have checked the descriptors as is + PCA, in [this version](https://www.kaggle.com/code/antoninadolgorukova/op2-bl-submission?scriptVersionId=145674979) of my baseline notebook - LB score is much worse than 0s (0.675).\n\nIt confirms what I said, feature selection is needed here, followed by some standartization/normalization.", "2474411": "Thank you for the introductory webinar", "2480625": "Thank you, very much! Alexander!", "2483559": "šØāš¬ Brainstorm on \"Kaggle: Open Problems - Single-Cell Perturbations.\"\nāļø Tuesday 17 October, 17.30 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231017T153000Z%2F20231017T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20%20%20Brainstorm%20on%20%20%22Kaggle%3A%20Open%20Problems%20-%20Single-Cell%20Perturbations.%22)\n\n Let us discuss what's up on Kaggle challenge - Open Problems - Single-Cell Perturbations: overview proposed public solutions, CV-schemes, features constructions, some insights from biological data and so on. It is planned to be pretty informal - a kind of discussion and opinions exchange - everybody welcome. Some notes can be found [link1](https://docs.google.com/presentation/d/1wiz0Wmt4D54pqMMsIOyJHuQYMZ3hTBZQQnjbLzwoGYY/edit?usp=sharing), [link2] (https://docs.google.com/spreadsheets/d/1APN63PMaWZygVjYimK9Ivt0RvifdAU5JRYkxiDn4szw/edit?usp=sharing), [link3](https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/445883).\n\nZoom link will be available here 5 minutes before the start. \nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga", "2485975": "Zoom link:\nhttps://us02web.zoom.us/j/81679339241?pwd=M0E2OWk0ZmJ6S2xLNGZoWlVJNEYvdz09", "2489501": "alexandervc Are these meetings still happening?", "2489538": "There were two , hopefully there will be more , announces will be here", "2512754": "šØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"\nāļø Monday 06 November, 17.00 (CET time)\n\n[Add to google calendat](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231106T160000Z%2F20231106T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=%20Antonina%20Dolgorukova%20%22Single%20Cell%20Perturbations%20data%20analysis%22)\n\nAntonina will share outcomes of here analysis based on the following notebook:\nhttps://www.kaggle.com/code/antoninadolgorukova/op2-adata-analysis-with-seurat\n\nZoom link will be available here 5 minutes before the start.\nVideo record will be available on the youtube channel soon afterwards: https://www.youtube.com/c/SciBerloga", "2514918": "Zoom link: https://us02web.zoom.us/j/89080226552?pwd=aGc1NElmWlVuZEljYjBaS1pZaUN2QT09\n\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"\nāļø Monday 06 November, 17.00 (CET time)", "2515295": "š¹ Video: https://youtu.be/lcc5vY-Pycs?si=NDz8KEarMFN8Llka\n\nšØāš¬ Antonina Dolgorukova \"Single Cell Perturbations data analysis with Seurat\"\nāļø Monday 06 November, 17.00 (CET time)", "2518538": "Thanks for the webinars @alexandervc - sounds like a quick intro to the competition for people who are late enough :))", "2553491": "Pyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it: \n\nšØāš¬ Anton Vakhrushev \"SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"\nāļø Monday 11 December 17.00 (CET time)\n\n[Add to Google Calendar](https://calendar.google.com/calendar/render?action=TEMPLATE&dates=20231211T160000Z%2F20231211T173000Z&details=Zoom%20link%20will%20be%20available%20in%20https%3A%2F%2Ft.me%2Fsberlogabig%20%20shortly%20before%20start%20of%20the%20talk.&text=Anton%20Vakhrushev%20%22SketchBoost%3A%20Fast%20Gradient%20Boosted%20Decision%20Tree%20for%20Multioutput%20Problems%22)\n\nTwitter: https://x.com/sberloga/status/1733589792024080479?s=20\nWould be happy for retwit", "2557675": "Zoom link:\nhttps://us02web.zoom.us/j/86106703038?pwd=VFBtUm9HTW5wWmFUUTJnUG1scytOQT09", "2558549": "š¹ Video: https://youtu.be/5xRxuDh_cGk\nš Presentation: https://t.me/sberlogacompete/10211\n\nPyboost has been used by many during the competition. Happy to invite everybody to a webinar where its creator Anton Vakhrushev will explain more on it:\n\nšØāš¬ Anton Vakhrushev, Leonid Iosipoi \"Pyboost/SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems\"\nāļø Monday 11 December 17.00 (CET time)\n\nTwitter: https://x.com/sberloga/status/1733589792024080479?s=20", "2994862": "nice work!" }, "source": "meta" }