What can we learn from Medical warning systems to help improve this Earthquake warning system?
\n\nAs a potential lifesaving warning device, LANL Earthquake Prediction has much in common with other lifesaving warning devices, most notably medical devices. There is additional commonality, in that many data scientists and researchers work on several machine learning areas, so I expect participants analyzing geophysical data in this competition may also be interested in biomedical signal analysis.
\n\nBecause of this, I wanted to share the very latest \"Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML) – Based Software as a Medical Device (SaMD)\" published earlier this month by the United States Food and Drug Administration (FDA).
\n\nThe FDA Document helps us categorize the system, describe ongoing changes to the system, and discusses good machine learning practices.
\n\nI will break down the components of the proposed regulatory framework, and use this example Kernel to demonstrate how it might impact your future work as a researcher, programmer, and data scientist.
\n\nhttps://www.kaggle.com/pnussbaum/earthquake-pred-cnn-medical-analogy-v07
\n\nPlease up-vote the Kernel if you find this useful, and good luck on this competition and your future careers (medical or otherwise)!
\n\n---Paul
", "messages": [ { "id": "521774", "postDate": "04/23/2019 11:58:06", "content": "What can we learn from Medical warning systems to help improve this Earthquake warning system?
\n\nAs a potential lifesaving warning device, LANL Earthquake Prediction has much in common with other lifesaving warning devices, most notably medical devices. There is additional commonality, in that many data scientists and researchers work on several machine learning areas, so I expect participants analyzing geophysical data in this competition may also be interested in biomedical signal analysis.
\n\nBecause of this, I wanted to share the very latest \"Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML) – Based Software as a Medical Device (SaMD)\" published earlier this month by the United States Food and Drug Administration (FDA).
\n\nThe FDA Document helps us categorize the system, describe ongoing changes to the system, and discusses good machine learning practices.
\n\nI will break down the components of the proposed regulatory framework, and use this example Kernel to demonstrate how it might impact your future work as a researcher, programmer, and data scientist.
\n\nhttps://www.kaggle.com/pnussbaum/earthquake-pred-cnn-medical-analogy-v07
\n\nPlease up-vote the Kernel if you find this useful, and good luck on this competition and your future careers (medical or otherwise)!
\n\n---Paul
", "rawMarkdown": "What can we learn from Medical warning systems to help improve this Earthquake warning system?\n\nAs a potential lifesaving warning device, LANL Earthquake Prediction has much in common with other lifesaving warning devices, most notably medical devices. There is additional commonality, in that many data scientists and researchers work on several machine learning areas, so I expect participants analyzing geophysical data in this competition may also be interested in biomedical signal analysis. \n\nBecause of this, I wanted to share the very latest \"Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML) – Based Software as a Medical Device (SaMD)\" published earlier this month by the United States Food and Drug Administration (FDA).\n\nThe FDA Document helps us categorize the system, describe ongoing changes to the system, and discusses good machine learning practices.\n\nI will break down the components of the proposed regulatory framework, and use this example Kernel to demonstrate how it might impact your future work as a researcher, programmer, and data scientist.\n\nhttps://www.kaggle.com/pnussbaum/earthquake-pred-cnn-medical-analogy-v07 \n\nPlease up-vote the Kernel if you find this useful, and good luck on this competition and your future careers (medical or otherwise)!\n\n---Paul", "votes": null }, { "id": "540088", "postDate": "05/31/2019 02:29:02", "content": "Apologies for the late posting.
\n\nI finished a Kernel for you to look over.
\n\nThis Kernel demonstrates a \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.
\n\nA \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\"}]
\n\nWhen using a k-means cross validation scheme, we take a portion of the training set and save it for validation. The number of times this is repeated is “k”. When using a traditional “Leave One Out” (LOO) k-means cross-validation, we are taking the most rigorous approach. We leave one of the experiments (one of the people in a an experiment involving humans, or one earthquake experiment here). This is the most rigorous because we are not seeding the training set with “parts of that experiment” – since in the real world, we won’t get that luxury. The modification I made in this Kernel is to pair up experiments (so “Leave Two Out”) yielding a k value of 8. So it is an 8-fold cross validation with LTO experiments. Each of these creates a solution neural network, and together they form an ensemble of 8 solutions, whose results are averaged as an ensemble. Note: Each solution is deleted before calculating the next, to save memory in Kaggle.
\n\nThis Kernel also demonstrates a \"zero feature extraction\" method to solve the problem. Keeps trainable parameter count very low (less than 32,000 trainable parameters solve the whole problem)
\n\nGets a pretty good LB score, with no feature extraction. The CNN does all the feature extraction itself, using Discrete Wavelet Transform. This is accomplished using a novel method of \"pseudo-residual\" to calculate detail coefficients.
\n\nhttps://www.kaggle.com/pnussbaum/dwt-earthquake-w-lto-v01
\n\nEnjoy, and good luck!
", "rawMarkdown": "Apologies for the late posting.\n\nI finished a Kernel for you to look over.\n\nThis Kernel demonstrates a \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\n\nA \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\"}]\n\nWhen using a k-means cross validation scheme, we take a portion of the training set and save it for validation. The number of times this is repeated is “k”. When using a traditional “Leave One Out” (LOO) k-means cross-validation, we are taking the most rigorous approach. We leave one of the experiments (one of the people in a an experiment involving humans, or one earthquake experiment here). This is the most rigorous because we are not seeding the training set with “parts of that experiment” – since in the real world, we won’t get that luxury. The modification I made in this Kernel is to pair up experiments (so “Leave Two Out”) yielding a k value of 8. So it is an 8-fold cross validation with LTO experiments. Each of these creates a solution neural network, and together they form an ensemble of 8 solutions, whose results are averaged as an ensemble. Note: Each solution is deleted before calculating the next, to save memory in Kaggle.\n\nThis Kernel also demonstrates a \"zero feature extraction\" method to solve the problem. Keeps trainable parameter count very low (less than 32,000 trainable parameters solve the whole problem)\n\nGets a pretty good LB score, with no feature extraction. The CNN does all the feature extraction itself, using Discrete Wavelet Transform. This is accomplished using a novel method of \"pseudo-residual\" to calculate detail coefficients.\n\nhttps://www.kaggle.com/pnussbaum/dwt-earthquake-w-lto-v01\n\nEnjoy, and good luck!", "votes": null } ], "comments": [ { "id": 540088, "author_name": "pnussbaum", "author_url": "", "post_date": "05/31/2019 02:29:02", "content": "Apologies for the late posting.
\n\nI finished a Kernel for you to look over.
\n\nThis Kernel demonstrates a \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.
\n\nA \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\"}]
\n\nWhen using a k-means cross validation scheme, we take a portion of the training set and save it for validation. The number of times this is repeated is “k”. When using a traditional “Leave One Out” (LOO) k-means cross-validation, we are taking the most rigorous approach. We leave one of the experiments (one of the people in a an experiment involving humans, or one earthquake experiment here). This is the most rigorous because we are not seeding the training set with “parts of that experiment” – since in the real world, we won’t get that luxury. The modification I made in this Kernel is to pair up experiments (so “Leave Two Out”) yielding a k value of 8. So it is an 8-fold cross validation with LTO experiments. Each of these creates a solution neural network, and together they form an ensemble of 8 solutions, whose results are averaged as an ensemble. Note: Each solution is deleted before calculating the next, to save memory in Kaggle.
\n\nThis Kernel also demonstrates a \"zero feature extraction\" method to solve the problem. Keeps trainable parameter count very low (less than 32,000 trainable parameters solve the whole problem)
\n\nGets a pretty good LB score, with no feature extraction. The CNN does all the feature extraction itself, using Discrete Wavelet Transform. This is accomplished using a novel method of \"pseudo-residual\" to calculate detail coefficients.
\n\nhttps://www.kaggle.com/pnussbaum/dwt-earthquake-w-lto-v01
\n\nEnjoy, and good luck!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "521774": "What can we learn from Medical warning systems to help improve this Earthquake warning system?\n\nAs a potential lifesaving warning device, LANL Earthquake Prediction has much in common with other lifesaving warning devices, most notably medical devices. There is additional commonality, in that many data scientists and researchers work on several machine learning areas, so I expect participants analyzing geophysical data in this competition may also be interested in biomedical signal analysis. \n\nBecause of this, I wanted to share the very latest \"Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML) – Based Software as a Medical Device (SaMD)\" published earlier this month by the United States Food and Drug Administration (FDA).\n\nThe FDA Document helps us categorize the system, describe ongoing changes to the system, and discusses good machine learning practices.\n\nI will break down the components of the proposed regulatory framework, and use this example Kernel to demonstrate how it might impact your future work as a researcher, programmer, and data scientist.\n\nhttps://www.kaggle.com/pnussbaum/earthquake-pred-cnn-medical-analogy-v07 \n\nPlease up-vote the Kernel if you find this useful, and good luck on this competition and your future careers (medical or otherwise)!\n\n---Paul", "540088": "Apologies for the late posting.\n\nI finished a Kernel for you to look over.\n\nThis Kernel demonstrates a \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\n\nA \"Leave Two Out\" k-means cross-validation scheme to insure accurate results.\"}]\n\nWhen using a k-means cross validation scheme, we take a portion of the training set and save it for validation. The number of times this is repeated is “k”. When using a traditional “Leave One Out” (LOO) k-means cross-validation, we are taking the most rigorous approach. We leave one of the experiments (one of the people in a an experiment involving humans, or one earthquake experiment here). This is the most rigorous because we are not seeding the training set with “parts of that experiment” – since in the real world, we won’t get that luxury. The modification I made in this Kernel is to pair up experiments (so “Leave Two Out”) yielding a k value of 8. So it is an 8-fold cross validation with LTO experiments. Each of these creates a solution neural network, and together they form an ensemble of 8 solutions, whose results are averaged as an ensemble. Note: Each solution is deleted before calculating the next, to save memory in Kaggle.\n\nThis Kernel also demonstrates a \"zero feature extraction\" method to solve the problem. Keeps trainable parameter count very low (less than 32,000 trainable parameters solve the whole problem)\n\nGets a pretty good LB score, with no feature extraction. The CNN does all the feature extraction itself, using Discrete Wavelet Transform. This is accomplished using a novel method of \"pseudo-residual\" to calculate detail coefficients.\n\nhttps://www.kaggle.com/pnussbaum/dwt-earthquake-w-lto-v01\n\nEnjoy, and good luck!" }, "source": "meta" }