Interpretable AI: Not Just For Regulators

下载 1

Spark开源社区

发布于

8413

人观看

#信息技术

Machine learning systems are used today to make life-altering decisions about employment, bail, parole, and lending. Moreover, the scope of decisions delegated to machine learning systems seems likely only to expand in the future. Unfortunately serious discrimination, privacy, and even accuracy concerns can be raised about these systems. Many researchers and practitioners are tackling disparate impact, inaccuracy, privacy issues, and security problems with a number of brilliant, but often siloed, approaches. This presentation illustrates how to combine innovations from several sub-disciplines of machine learning research to train explainable, fair, trustable, and accurate predictive modeling systems. Together these techniques create a new and truly human-centered type of machine learning suitable for use in business- and life-critical decision support.

展开查看详情

1 .WIFI SSID:SparkAISummit | Password: UnifiedAnalytics

2 .Beyond Reason Codes A Blueprint for Human-Centered, Low-Risk ML Patrick Hall, H2O.ai #UnifiedAnalytics #SparkAISummit

3 . Contents • Blueprint • EDA • Benchmark • Training • Post-Hoc Analysis • Review • Deployment • Appeal • Iterate • Questions #UnifiedAnalytics #SparkAISummit 3

4 . Blueprint #UnifiedAnalytics #SparkAISummit 4

5 .EDA and Data Visualization • OSS: H2O-3 Aggregator • References: Visualizing Big Data Outliers through Distributed Aggregation; The Grammar of Graphics #UnifiedAnalytics #SparkAISummit 5

6 . Establish Benchmarks Establishing a benchmark from which to gauge improvements in accuracy, fairness, interpretability, privacy, or security is crucial for good (“data”) science and for compliance. #UnifiedAnalytics #SparkAISummit 6

7 .Manual, Private, Sparse or Straightforward Feature Engineering • OSS: Pandas Profiler, Feature Tools • References: Deep Feature Synthesis: Towards Automating Data Science Endeavors; Label, Segment, Featurize: A Cross Domain Framework for Prediction Engineering #UnifiedAnalytics #SparkAISummit 7

8 .Preprocessing for Fairness, Privacy or Security • OSS: IBM AI360 • References: Data Preprocessing Techniques for Classification Without Discrimination; Certifying and Removing Disparate Impact; Optimized Pre-processing for Discrimination Prevention; Privacy- Preserving Data Mining #UnifiedAnalytics #SparkAISummit 8

9 .Constrained, Fair, Interpretable, Private or Simple Models References: Explainable Neural Networks Based on Additive Index Models (XNN); Learning Fair Representations (LFR); Locally Interpretable Models and Effects Based on Supervised Partitioning (LIME-SUP); Private Aggregation of Teacher Ensembles (PATE); Scalable Bayesian Rule Lists (SBRL) #UnifiedAnalytics #SparkAISummit 9

10 .Traditional Model Assessment and Diagnostics • Residual analysis, Q-Q plots, AUC and lift curves confirm model is accurate and meets assumption criteria. • Implemented as model diagnostics in Driverless AI. #UnifiedAnalytics #SparkAISummit 10

11 .Post-hoc Explanations • OSS: lime, shap • References: Why Should I Trust You?: Explaining the Predictions of Any Classifier; A Unified Approach to Interpreting Model Predictions; Please Stop Explaining Black Box Models for High Stakes Decisions (criticism) #UnifiedAnalytics #SparkAISummit 11

12 . Interlude: The Time–Tested Shapley Value 1. In the beginning: A Value for N-Person Games, 1953 2. Nobel-worthy contributions: The Shapley Value: Essays in Honor of Lloyd S. Shapley, 1988 3. Shapley regression: Analysis of Regression in Game Theory Approach, 2001 4. First reference in ML? Fair Attribution of Functional Contribution in Artificial and Biological Networks, 2004 5. Into the ML research mainstream, i.e. JMLR: An Efficient Explanation of Individual Classifications Using Game Theory, 2010 6. Into the real-world data mining workflow ... finally: Consistent Individualized Feature Attribution for Tree Ensembles, 2017 7. Unification: A Unified Approach to Interpreting Model Predictions, 2017 #UnifiedAnalytics #SparkAISummit 12

13 .Model Debugging for Accuracy, Privacy or Security OSS: cleverhans, pdpbox, what-if tool References: A Marauder’s Map of Security and Privacy in Machine Learning: An overview of current and future research directions for making machine learning secure and private; Modeltracker: Redesigning Performance Analysis Tools for Machine Learning; The Security of Machine Learning #UnifiedAnalytics #SparkAISummit 13

14 .Post-hoc Disparate Impact Assessment and Remediation OSS: aequitas, IBM AI360 , themis References: Equality of Opportunity in Supervised Learning; Certifying and Removing Disparate Impact #UnifiedAnalytics #SparkAISummit 14

15 .Human Review and Documentation References: Model Cards for Model Reporting #UnifiedAnalytics #SparkAISummit 15

16 .Deployment, Management and Monitoring OSS: mlflow, modeldb, awesome- machine-learning-ops metalist Reference: Model DB: A System for Machine Learning Model Management #UnifiedAnalytics #SparkAISummit 16

17 . Human Appeal Very important, may require custom implementation for each deployment environment? #UnifiedAnalytics #SparkAISummit 17

18 .Iterate: Use Gained Knowledge to Improve Accuracy, Fairness, Interpretability, Privacy or Security Improvements, KPIs should not be restricted to accuracy alone. #UnifiedAnalytics #SparkAISummit 18

19 . Open Conceptual Questions • How much automation is appropriate, 100%? • How to automate learning by iteration, reinforcement learning? • How to implement human appeals, is it productizable? Mobile Apps? #UnifiedAnalytics #SparkAISummit 19

20 . Resources In-Depth Open Source Interpretability Technique Examples: https://github.com/jphall663/interpretable_machine_learning_with_python "Awesome" Machine Learning Interpretability Resource List: https://github.com/jphall663/awesome-machine-learning-interpretability #UnifiedAnalytics #SparkAISummit 20

21 . References Agrawal, Rakesh and Ramakrishnan Srikant (2000). “Privacy-Preserving Data Mining.” In: ACM Sigmod Record. Vol. 29. 2. URL: http://alme1.almaden.ibm.com/cs/projects/iis/hdb/Publications/papers/sigmod00_privacy.pdf. ACM, pp. 439–450. Amershi, Saleema et al. (2015). “Modeltracker: Redesigning Performance Analysis Tools for Machine Learning.” In: Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems. URL: https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/amershi.CHI2015.ModelTracker.pdf. ACM, pp. 337–346. Barreno, Marco et al. (2010). “The Security of Machine Learning.” In: Machine Learning 81.2. URL: https://people.eecs.berkeley.edu/~adj/publications/paper-files/SecML-MLJ2010.pdf, pp. 121–148. Calmon, Flavio et al. (2017). “Optimized Pre-processing for Discrimination Prevention.” In: Advances in Neural Information Processing Systems. URL: http://papers.nips.cc/paper/6988-optimized-pre-processing-for-discrimination- prevention.pdf, pp. 3992–4001. Feldman, Michael et al. (2015). “Certifying and Removing Disparate Impact.” In: Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. URL: https://arxiv.org/pdf/1412.3756.pdf. ACM, pp. 259–268. #UnifiedAnalytics #SparkAISummit 21

22 . References Hardt, Moritz, Eric Price, Nati Srebro, et al. (2016). “Equality of Opportunity in Supervised Learning.” In: Advances in neural information processing systems. URL: http://papers.nips.cc/paper/6374-equality-of-opportunity-in-supervised-learning.pdf, pp. 3315–3323. Hu, Linwei et al. (2018). “Locally Interpretable Models and Effects Based on Supervised Partitioning (LIME-SUP).” In: arXiv preprint arXiv:1806.00663. URL: https://arxiv.org/ftp/arxiv/papers/1806/1806.00663.pdf. Kamiran, Faisal and Toon Calders (2012). “Data Preprocessing Techniques for Classification Without Discrimination.” In: Knowledge and Information Systems 33.1. URL: https://link.springer.com/content/pdf/10.1007/s10115-011-0463- 8.pdf, pp. 1–33. Kanter, James Max, Owen Gillespie, and Kalyan Veeramachaneni (2016). “Label, Segment, Featurize: A Cross Domain Framework for Prediction Engineering.” In: Data Science and Advanced Analytics (DSAA), 2016 IEEE International Conference on. URL: http://www.jmaxkanter.com/static/papers/DSAA_LSF_2016.pdf. IEEE, pp. 430–439. Kanter, James Max and Kalyan Veeramachaneni (2015). “Deep Feature Synthesis: Towards Automating Data Science Endeavors.” In: Data Science and Advanced Analytics (DSAA), 2015. 36678 2015. IEEE International Conference on. URL: https://groups.csail.mit.edu/EVO-DesignOpt/groupWebSite/uploads/Site/DSAA_DSM_2015.pdf. IEEE, pp. 1–10. #UnifiedAnalytics #SparkAISummit 22

23 . References Keinan, Alon et al. (2004). “Fair Attribution of Functional Contribution in Artificial and Biological Networks.” In: Neural Computation 16.9. URL: https://www.researchgate.net/profile/Isaac_Meilijson/ publication/2474580_Fair_Attribution_of_Functional_Contribution_in_Artificial_and_ Biological_Networks/links/09e415146df8289373000000/Fair-Attribution-of-Functional-Contribution-in-Artificial-and- Biolo, pp. 1887–1915. Kononenko, Igor et al. (2010). “An Efficient Explanation of Individual Classifications Using Game Theory.” In: Journal of Machine Learning Research 11.Jan. URL: http://www.jmlr.org/papers/volume11/strumbelj10a/strumbelj10a.pdf, pp. 1– 18. Lipovetsky, Stan and Michael Conklin (2001). “Analysis of Regression in Game Theory Approach.” In: Applied Stochastic Models in Business and Industry 17.4, pp. 319–330. Lundberg, Scott M., Gabriel G. Erion, and Su-In Lee (2017). “Consistent Individualized Feature Attribution for Tree Ensembles.” In: Proceedings of the 2017 ICML Workshop on Human Interpretability in Machine Learning (WHI 2017). Ed. by Been Kim et al. URL: https://openreview.net/pdf?id=ByTKSo-m-. ICML WHI 2017, pp. 15–21. Lundberg, Scott M and Su-In Lee (2017). “A Unified Approach to Interpreting Model Predictions.” In: Advances in Neural Information Processing Systems 30. Ed. by I. Guyon et al. URL: http://papers.nips.cc/paper/7062-a-unified-approach- to-interpreting-model-predictions.pdf. Curran Associates, Inc., pp. 4765–4774. #UnifiedAnalytics #SparkAISummit 23

24 . References Mitchell, Margaret et al. (2019). “Model Cards for Model Reporting.” In: Proceedings of the Conference on Fairness, Accountability, and Transparency. URL: https://arxiv.org/pdf/1810.03993.pdf. ACM, pp. 220–229. Papernot, Nicolas (2018). “A Marauder’s Map of Security and Privacy in Machine Learning: An overview of current and future research directions for making machine learning secure and private.” In: Proceedings of the 11th ACM Workshop on Artificial Intelligence and Security. URL: https://arxiv.org/pdf/1811.01134.pdf. ACM. Papernot, Nicolas et al. (2018). “Scalable Private Learning with PATE.” In: arXiv preprint arXiv:1802.08908. URL: https://arxiv.org/pdf/1802.08908.pdf. Ribeiro, Marco Tulio, Sameer Singh, and Carlos Guestrin (2016). “Why Should I Trust You?: Explaining the Predictions of Any Classifier.” In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. URL: http://www.kdd.org/kdd2016/papers/files/rfp0573-ribeiroA.pdf. ACM, pp. 1135–1144. Rudin, Cynthia (2018). “Please Stop Explaining Black Box Models for High Stakes Decisions.” In: arXiv preprint arXiv:1811.10154. URL: https://arxiv.org/pdf/1811.10154.pdf. Shapley, Lloyd S (1953). “A Value for N-Person Games.” In: Contributions to the Theory of Games 2.28. URL: http://www.library.fa.ru/files/Roth2.pdf#page=39, pp. 307–317. #UnifiedAnalytics #SparkAISummit 24

25 . References Shapley, Lloyd S, Alvin E Roth, et al. (1988). The Shapley Value: Essays in Honor of Lloyd S. Shapley. URL: http://www.library.fa.ru/files/Roth2.pdf. Cambridge University Press. Vartak, Manasi et al. (2016). “Model DB: A System for Machine Learning Model Management.” In: Proceedings of the Workshop on Human-In-the-Loop Data Analytics. URL: https://www- cs.stanford.edu/~matei/papers/2016/hilda_modeldb.pdf. ACM, p. 14. Vaughan, Joel et al. (2018). “Explainable Neural Networks Based on Additive Index Models.” In: arXiv preprint arXiv:1806.01933. URL: https://arxiv.org/pdf/1806.01933.pdf. Wilkinson, Leland (2006). The Grammar of Graphics. Wilkinson, Leland (2018). “Visualizing Big Data Outliers through Distributed Aggregation.” In: IEEE Transactions on Visualization & Computer Graphics. URL: https://www.cs.uic.edu/~wilkinson/Publications/outliers.pdf. Yang, Hongyu, Cynthia Rudin, and Margo Seltzer (2017). “Scalable Bayesian Rule Lists.” In: Proceedings of the 34th International Conference on Machine Learning (ICML). URL: https://arxiv.org/pdf/1602.08610.pdf. Zemel, Rich et al. (2013). “Learning Fair Representations.” In: International Conference on Machine Learning. URL: http://proceedings.mlr.press/v28/zemel13.pdf, pp. 325–333. #UnifiedAnalytics #SparkAISummit 25

26 .DON’T FORGET TO RATE AND REVIEW THE SESSIONS SEARCH SPARK + AI SUMMIT

0点赞

0收藏

1下载