Spring 2021

Littlestone, N.; Warmuth, M. (1994). "The Weighted Majority Algorithm". Information and Computation 108: 212–261. doi:10.1006/inco.1994.1009
Yoav Freund and Robert E. Schapire. A decision-theoretic generalization of on-line learning and an application to boosting. Journal of Computer and System Sciences, 55(1):119--139,
S. Shalev-Shwartz Online Learning and Convex Optimization DOI: 10.1561/2200000018

Lecture 3: Introduction to Active Learning

Topics

Core concepts and terminology, Heuristic query selection strategies, part 1

Slides

Required Reading

Sections 1-3 of "Active Learning Literature Survey", B. Settles, UW Madison CS Tech Report 1648

Lecture 4: Heuristic query selection strategies

Topics

Heuristic query selection strategies, part 2

Slides

Required Reading

Sections 1-3 of "Active Learning Literature Survey", B. Settles, UW Madison CS Tech Report 1648

Lecture 5: Advanced query selection methods, part 1

Topics

The two faces of active learning
Hypothesis space search methods, part 1
- The CAL algorithm

Slides

Required Reading

S. Dasgupta, Two Faces of Active Learning doi:10.1016/j.tcs.2010.12.054

Optional Reading

CAL: Cohn, Atlas, Ladner; "Improving Generalization with Active Learning" Machine Learning May 1994, Volume 15, Issue 2, pp 201-221

Lecture 6: Advanced query selection methods, part 2

Topics

The A²algorithm
The DHM algorithm

Slides

Required Reading

"DHM": Dasgupta, Hsu, Monteleoni "A general agnostic active learning algorithm" NIPS '08
- Alternatively, read the description of the DHM algorithm in the "Two Faces of Active Learning" paper.

Optional Advanced Reading

"A²": Balcan, Beygelzimer, Langford "Agnostic Active Learning" J Computer and System Sciences Volume 75, Issue 1, January 2009, Pages 78–89

Lecture 7: Advanced query selection methods, part 3

Topics

Hypothesis space search methods, part 3
- The IWAL algorithm
Cluster exploitation methods, part 1
- The ZLG algorithm

Slides

Required Reading

"IWAL": Beygelzimer, Dasgupta, Langford, "Importance Weighted Active Learning" ICML 09
"ZLG": Zhu, Lafferty, Ghahramani, "Combining Active Learning and Semi-Supervised Learning Using Gaussian Fields and Harmonic Functions" ICML 03

Optional Reading

A longer version of the IWAL paper.

Lecture 8: Advanced query selection methods, part 4

Topics

Cluster exploitation methods, part 2
- Background & the DH algorithm
- The PLAL algorithm

Slides

Required Reading

"DH": Dasgupta and Hsu; "Hierarchical Sampling for Active Learning" ICML 08

Optional Advanced Reading

"PLAL": Urner, Wulff, Ben-David "PLAL: Cluster-based Active Learning" COLT 13

Lecture 9: A Theory of Active Learning

Topics

Computational Learning Theory
- PAC Learning
- VC dimension
Disagreement Coefficient

Slides

Optional Reading

Chapter 2 of Theoretical Foundations of Active Learning, Steve Hanneke, Ph.D. Dissertation, Machine Learning Department, Carnegie Mellon University CMU-ML-09-106
Castro and Nowak "Minimax Bounds for Active Learning" Information Theory, IEEE Transactions on (Volume:54 , Issue: 5 ) P. 2339 - 2353
V. Vapnik. Statistical Learning Theory. Wiley, 1998.
Balcan, Hanneke, Vaughn “The True Sample Complexity of Active Learning”, Machine Learning, 2010, Volume 80, Issue 2, pp 111-139

Lecture 10: Active Learning for regression, part 1

Topics

Active Learning for Non-parametric Regression
Optimal Design of Experiments (DOE)

Slides

Required Reading

Read sections 1 & 2 of Optimal Design
"RDP": Faster Rates in Regression via Active Learning, Castro, Willett, Nowak UW Madison Technical Report ECE-05-3

Optional Advanced Reading

A longer version of the RDP paper: "RDP": Faster Rates in Regression via Active Learning, Castro, Willett, Nowak UW Madison Technical Report ECE-05-3
Smith, Kirstine (1918). "On the Standard Deviations of Adjusted and Interpolated Values of an Observed Polynomial Function and its Constants and the Guidance They Give Towards a Proper Choice of the Distribution of the Observations". Biometrika 12 (1): 1–85. doi:10.2307/2331929
Design and Analysis of Experiments Douglas Montgomery, Wiley, 8th Edition
Convex Optimization Boyd and Vandenberghe
Robust Design of Biological Experiments Flaherty, Jordan, Arkin, NIPS, 2006

Lecture 11: Active Learning for regression, part 2

Topics

Active Learning for regression using DOE techniques

Slides

Required Reading

"ALICE" Active Learning in Approximately Linear Regression Based on Conditional Expectation of Generalization Error Sugiyama; JMLR 7(Jan):141--166, 2006
"LapRDD" Laplacian Regularized D-optimal Design for active learning and its application to image retrieval. He X, IEEE Trans Image Process. 2010 Jan;19(1):254-63. doi: 10.1109/TIP.2009.2032342

Optional Advanced Reading

M. Belkin, P. Niyogi, and V. Sindhwani, “Manifold regularization: A geometric framework for learning from labeled and unlabeled examples,” J. Mach. Learn.Res., vol. 7, pp. 2399–2434, 2006
Pool-based active learning in approximate linear regression Sugiyama & Nakajima, ML, vol.75, no.3, pp.249-274, 2009

Lecture 12: Proactive Learning; Active Feature-value selection

Topics

Proactive Learning
Active Feature Value Acquisition

Slides

Required Reading

Donmez, P., Carbonell, J.G.: Proactive Learning: Cost-Sensitive Active Learning with Multiple Imperfect Oracles, in Proceedings of the 17th ACM Conference on Information and Knowledge Management (CIKM '08), 2008.
Saar-Tsechansky, et al "Active Feature-Value Acquisition" Mgmt Sci, 2009

Optional Reading

Zhang and Chaudhuri, "Active Learning from Weak and Strong Labelers", NIPS 2015
A longer version of the Zhang and Chaudhuri, paper that includes appendices.
Learning from Weak Teachers Urner, Ben-David, and Shamir, AISTATS, 2012, pages 1252-1260

Lecture 13: Sequential Optimization; Multi-Armed Bandits

Topics

Scientific Discovery vs Engineering Design
Sequential Model-Based Optimization & Multi-Armed Bandits

Slides

Required Reading

Bergstra et al "Algorithms for hyper-parameter optimization", NIPS, 2011

Optional Reading

Bayesian_optimization
Hyperparameter_optimization
A tutorial on Bayesian Optimization, P. Frazier
Practical Bayesian Optimization of Machine Learning Algorithms, Snoek et al, NIPS 2012, pp. 2951–2959

Optional Advanced Reading

Peter Auer, Nicolò Cesa-Bianchi, Paul Fischer "Finite-time Analysis of the Multiarmed Bandit Problem" Machine Learning May 2002, Volume 47, Issue 2, pp 235-256

Reading for fun

Reid, et al "Decision-making without a brain: how an amoeboid organism solves the two-armed bandit", 2016.DOI: 10.1098/rsif.2016.0030
- Associated videos

Lecture 14: Active Learning for Drug Screening

Topics

Cloud Laboratories
Active Learning for Drug Screening

Slides

Required Reading

Warmuth et al "Active Learning with Support Vector Machines in the Drug Discovery Process", J Chem Inf Comput Sci. 2003 Mar-Apr;43(2):667-73

Lecture 15: Automated Phenotyping

Topics

Phenotyping
Protein localization

Slides

Required Reading

Smith and Horvath, "Active Learning Strategies for Phenotypic Profiling of High-Content Screens", J Biomol Screen June 2014 vol. 19 no. 5 685-695

Optional Reading

Jarvik, et al "CD-tagging: a new approach to gene and protein discovery and analysis", Biotechniques. 1996 May;20(5):896-904
Sigal, et al "Generation of a fluorescently labeled endogenous protein library in living human cells" Nature Protocols 2, - 1515 - 1527 (2007)
Naik et al "Active machine learning-driven experimentation to determine compound effects on protein patterns",2016, eLife DOI: 10.7554/eLife.10047.001

Lecture 16: Bayesian Active Learning

Topics

Bayesian Estimation & Ordinary Differential Equation Models
Bayesian Active Learning

Slides

Required Reading

A Bayesian active learning strategy for sequential experimental design in systems biology, Pauwels, et al. BMC Syst Biol, 2014; 8(1): 102

Optional Reading

Kinetics of Influenza A Virus Infection in Humans, Baccam, et al JOURNAL OF VIROLOGY, 2006, p. 7590–7599
More detailed notes on Bayesian Estimation
More detailed notes on Bayesian Estimation 2

Lecture 17: Automated discovery of gene function

Topics

Logical Inference
Logic Based Active Learning

Slides

Required Reading

King et al "Functional genomic hypothesis generation and experimentation by a robot scientist" , Nature 427, 247-252, 2004. If you are off campus, you can also get a copy of the paper here.
- also see supplemental information

Optional Reading

King et al "The Automation of Science", Science 2009: Vol. 324 no. 5923 pp. 85-89
- also see supplemental information
Bryant et al Combining Inductive Logic Programming Active Learning and Robotics to Discover the Function of Genes, Electronic Transactions on Artificial Intelligence, Vol. 5 (2001), Section B, pp. 1-36
Sparkes et al "Towards Robot Scientists for autonomous scientific discovery", Autom Exp. 2010; 2: 1

Lecture 18: Automated learning of regulatory networks, part 1

Topics

Bayesian Networks
Today's paper

Slides

Required Reading

Introduction to Bayesian Networks
Cho et al "Reconstructing Causal Biological Networks through Active Learning", PLoS One, 2016 11(3): e0150611. doi:10.1371/journal.pone.0150611

Optional Reading

Bayesian Networks
Pournara and Wernisch "Reconstruction of gene networks using Bayesian learning and manipulation experiments" Bioinformatics. 2004 Nov 22;20(17):2934-42
Active Learning of Causal Bayes Net Structure KP Murphy, 2001
Ness et al "A Bayesian Active Learning Experimental Design for Inferring Signaling Networks" RECOMB 2017: Research in Computational Molecular Biology pp 134-156

Lecture 19: Automated learning of regulatory networks, part 2

Topics

Active Learning of Boolean Networks

Slides

Required Reading

Atias et al Experimental design schemes for learning Boolean network models, Bioinformatics. 2014 Sep 1; 30(17): i445–i452.

Lecture 20: Automated Protein Design, part 1

Topics

Active Learning for Protein Design

Slides

Required Reading

Danziger et al "Predicting Positive p53 Cancer Rescue Regions Using Most Informative Positive (MIP) Active Learning", PLoS Comp Bio 2009, DOI: 10.1371/journal.pcbi.1000498

Lecture 21: Automated Protein Design, part 2

Topics

Machine-Learning Guided Directed Evolution

Slides

Required Reading

Machine learning-assisted directed protein evolution with combinatorial libraries, Wu et al PNAS 2019, https://doi.org/10.1073/pnas.1901979116

Optional Reading

Fold Family-Regularized Bayesian Optimization for Directed Protein Evolution. Frisby and Langmead, 20th International Workshop on Algorithms in Bioinformatics (WABI) 2020, pages 1-18.

Lecture 22: Automated Protein Design, part 3

Topics

Reinforcement learning for biological sequence design

Slides

Required Reading

Reinforcement Learning
MODEL-BASED REINFORCEMENT LEARNING FOR BIOLOGICAL SEQUENCE DESIGN, Angermueller et al ICLR 2020

Lecture 23: Automation in Chemistry

Topics

Deep Reinforcement Learning for Optimizing Chemical Reactions

Slides

Required Reading

Optimizing Chemical Reactions with Deep Reinforcement Learning, Zhou et al, ACS Cent. Sci. 2017, 3, 12, 1337-1344 https://doi.org/10.1021/acscentsci.7b00492

Optional Advanced Reading

Tuning the molecular weight distribution from atom transfer radical polymerization using deep reinforcement learning, Li et al, MSDE, 2018 DOI: 10.1039/C7ME00131B
Controlling an organic synthesis robot with machine learning to search for new reactivity, Granda et al, Nature, 559, 377–381 (2018) doi:10.1038/s41586-018-0307-8

Lecture 24: Discovering natural laws, part 1

Topics

Symbolic Regression for learning natural laws

Slides

Required Reading

Distilling Free-Form Natural Laws from Experimental Data, Schmidt and Lipson, Science, 2009 Vol. 324, Issue 5923, pp. 81-85 DOI: 10.1126/science.1165893

Lecture 25: Discovering natural laws, part 2

Topics

Abstract Boolean Networks and Formal Reasoning

Slides

Required Reading

A method to identify and analyze biological programs through automated reasoning, Yordanov, et al npj Syst Biol Appl 2, 16010 (2016) doi:10.1038/npjsba.2016.10

Lecture 26: Discovering natural laws, part 3; Course summary

Topics

Discovering PDEs
Course Summary

Slides

Required Reading

Data-driven discovery of partial differential equations, Rudy, et al Science Advances 2017 DOI: 10.1126/sciadv.1602614

Page updated

Google Sites

Report abuse