/tsgan

Time-series Generative Adversarial Networks (fork from the ML-AIM research group on bitbucket))

Primary LanguagePythonOtherNOASSERTION

ML-AIM: Machine Learning and Artificial Intelligence for Medicine

Finance Quant Machine Learning

In this fork, I present some of my own adjustments to the original work. I have also developed a conditional MTS gan, here.

This repository contains the implementations of algorithms developed by the ML-AIM Laboratory.

  1. AutoPrognosis: Automated Clinical Prognostic Modeling, ICML 2018 software
  2. GAIN: a GAN based missing data imputation algorithm, ICML 2018 software
  3. INVASE: an Actor-critic model based instance wise feature selection algorithm, ICLR 2019 software
  4. GANITE: a GAN based algorithm for estimating individualized treatment effects, ICLR 2018 software
  5. DeepHit: a Deep Learning Approach to Survival Analysis with Competing Risks, AAAI 2018 software
  6. PATE-GAN: Generating Synthetic Data with Differential Privacy Guarantees, ICLR 2019 software
  7. KnockoffGAN: generating knockoffs for feature selection using generative adversarial networks, ICLR 2019 software
  8. Causal Multi-task Gaussian Processes: Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes, NIPS 2017 software
  9. Limits of Estimating Heterogeneous Treatment Effects:Guidelines for Practical Algorithm Designsoftware
  10. ASAC: Active Sensing using Actor-Critic Models, MLHC 2019 software
  11. DGPSurvival: Deep Multi-task Gaussian Processes for Survival Analysis with Competing Risks, NIPS 2018 software
  12. Symbolic Metamodeling Demystifying Black-box Models with Symbolic Metamodels, NeurIPS 2019 software
  13. DPBAG Differentially Private Bagging: Improved utility and cheaper privacy than subsample-and-aggregate, NeurIPS 2019 software
  14. TimeGAN Time-series Generative Adversarial Networks, NeurIPS 2019 software
  15. Attentiveness Attentive State-Space Modeling of Disease Progression, NeurIPS 2019 software
  16. GCIT: Conditional Independence Testing with Generative Adversarial Networks, NeurIPS 2019 software
  17. Counterfactual Recurrent Network: Estimating counterfactual treatment outcomes over time through adversarially balanced representations, ICLR 2020 software
  18. C3T Budget: Contextual constrained learning for dose-finding clinical trials, AISTATS 2020 software
  19. DKLITE: Learning Overlapping Representations for the Estimation of Individualized Treatment Effects, AISTATS 2020 software
  20. Dynamic disease network ddp: Learning Dynamic and Personalized Comorbidity Networks from Event Data using Deep Diffusion Processes, AISTATS 2020 software
  21. SMS-DKL: Stepwise Model Selection for Sequence Prediction via Deep Kernel Learning, AISTATS 2020 software

Prepared for release and maintained by AvdSchaar

Please send comments and suggestions to mihaelaucla@gmail.com

Citations

Please cite the ML-AIM repository and or the applicable papers if you use the software.

License

Copyright 2019, 2020, ML-AIM

The ML-AIM software is released under the 3-Clause BSD license unless mentioned otherwise by the respective algorithms.

See doc/install.md for installation instructions

Tutorials and or examples

  • AutoPrognosis: -- alg/autoprognosis/tutorial_autoprognosis_api.ipynb -- alg/autoprognosis/tutorial_autoprognosis_cli.ipynb
  • GAIN: alg/gain/tutorial_gain.ipynb
  • INVASE: alg/invase/tutorial_invase.ipynb
  • GANITE: alg/ganite/tutorial_ganite.ipynb
  • PATE-GAN: alg/pategan/tutorial_pategan.ipynb
  • KnockoffGAN: alg/knockoffgan/tutorial_knockoffgan.ipynb
  • ASAC: alg/asac/tutorial_asac.ipynb
  • DGPSurvival: alg/dgp_survival/tutorial_dgp.ipynb
  • Symbolic Metamodeling: -- alg/symbolic_metamodeling/1-_Introduction_to_Meijer_G-functions.ipynb -- alg/symbolic_metamodeling/2-_Metamodeling_of_univariate_black-box_functions_using_Meijer_G-functions.ipynb -- alg/symbolic_metamodeling/3-_Building_Symbolic_Metamodels.ipynb
  • Differentially Private Bagging: alg/dpbag/DPBag_Tutorial.ipynb
  • Time-series Generative Adversarial Networks: alg/timegan/tutorial_timegan.ipynb
  • Attentive State-Space Modeling of Disease Progression: alg/attentivess/Tutorial_for_Attentive_State-space_Models.ipynb
  • Conditional Independence Testing with Generative Adversarial Networks: alg/gcit/tutorial_gcit.ipynb
  • DKLITE: alg/dklite/tutorial_dklite.ipynb
  • SMS-DKL: alg/smsdkl/test_smsdkl.py

You can find a presentation by Prof. van der Schaar describing AutoPrognosis here: https://www.youtube.com/watch?v=d1uEATa0qIo

Version history

  • version 1.7: February 27, 2020: SMS-DKL
  • version 1.6: February 24, 2020: DKLITE and dynamic disease network ddp
  • version 1.5: February 23, 2020: C3T Budget
  • version 1.4: February 3, 2020: Counterfactual Recurrent Network
  • version 1.3: December 7, 2019: Conditional Independence Testing with Generative Adversarial Networks
  • version 1.1: November 30, 2019: Attentive State-Space Modeling
  • version 1.0: November 4, 2019: Differentially Private Bagging and Time-series Generative Adversarial Networks
  • version 0.9: October 25, 2019: Symbolic Metamodeling
  • version 0.8: September 29, 2019: DGP Survival
  • version 0.7: September 20, 2019: ASAC
  • version 0.6: August 5, 2019: Causal Multi-task Gaussian Processes
  • version 0.5: July 24, 2019: KnockoffGAN
  • version 0.4: June 18, 2019: Deephit and PATE-GAN

References

  1. AutoPrognosis: Automated Clinical Prognostic Modeling via Bayesian Optimization with Structured Kernel Learning
  2. Prognostication and Risk Factors for Cystic Fibrosis via Automated Machine Learning
  3. Cardiovascular Disease Risk Prediction using Automated Machine Learning: A Prospective Study of 423,604 UK Biobank Participants
  4. GAIN: Missing Data Imputation using Generative Adversarial Nets
  5. INVASE: Instance-wise Variable Selection using Neural Networks
  6. GANITE: Estimation of Individualized Treatment Effects using Generative Adversarial Nets
  7. KnockoffGAN: generating knockoffs for feature selection using generative adversarial networks
  8. Bayesian Inference of Individualized Treatment Effects using Multi-task Gaussian Processes
  9. Limits of Estimating Heterogeneous Treatment Effects:Guidelines for Practical Algorithm Design
  10. ASAC Active Sensing using Actor-Critic Models
  11. DGPSurvival: Deep Multi-task Gaussian Processes for Survival Analysis with Competing Risks
  12. GCIT: Conditional Independence Testing with Generative Adversarial Networks
  13. Counterfactual Recurrent Network: Estimating counterfactual treatment outcomes over time through adversarially balanced representations
  14. C3T Budget: Contextual constrained learning for dose-finding clinical trials
  15. SMS-DKL: Stepwise Model Selection for Sequence Prediction via Deep Kernel Learning
  16. Dua, D. and Graff, C. (2019). UCI Machine Learning Repository. Irvine, CA: University of California, School of Information and Computer Science.
  17. Scikit-learn: Machine Learning in Python, Pedregosa et al., JMLR 12, pp. 2825-2830, 2011.
  18. TensorFlow: Large-scale machine learning on heterogeneous systems, 2015. Software available from tensorflow.org.
  19. GPyOpt: A Bayesian Optimization framework in python
  20. scikit-survival survival analysis built on top of scikit-learn
  21. 3-Clause BSD license