Timezone: »

Interpretable Machine Learning for Complex Systems
Andrew Wilson · Been Kim · William Herlands

Thu Dec 08 11:00 PM -- 09:30 AM (PST) @ AC Barcelona, Sagrada Familia
Event URL: https://sites.google.com/site/nips2016interpretml/ »

Complex machine learning models, such as deep neural networks, have recently achieved great predictive successes for visual object recognition, speech perception, language modelling, and information retrieval. These predictive successes are enabled by automatically learning expressive features from the data. Typically, these learned features are a priori unknown, difficult to engineer by hand, and hard to interpret. This workshop is about interpreting the structure and predictions of these complex models.

Interpreting the learned features and the outputs of complex systems allows us to more fundamentally understand our data and predictions, and to build more effective models. For example, we may build a complex model to predict long range crime activity. But by interpreting the learned structure of the model, we can gain new insights into the processing driving crime events, enabling us to develop more effective public policy. Moreover, if we learn, for example, that the model is making good predictions by discovering how the geometry of clusters of crime events affect future activity, we can use this knowledge to design even more successful predictive models.

This 1 day workshop is focused on interpretable methods for machine learning, with an emphasis on the ability to learn structure which provides new fundamental insights into the data, in addition to accurate predictions. We will consider a wide range of topics, including deep learning, kernel methods, tensor methods, generalized additive models, rule based models, symbolic regression, visual analytics, and causality. A poster session, coffee breaks, and a panel guided discussion will encourage interaction between attendees. We wish to carefully review and enumerate modern approaches to the challenges of interpretability, share insights into the underlying properties of popular machine learning algorithms, and discuss future directions.

Thu 11:45 p.m. - 12:00 a.m. [iCal]
Opening Remarks
Fri 12:00 a.m. - 12:30 a.m. [iCal]
Honglak Lee (Invited Talk)
Fri 12:30 a.m. - 1:00 a.m. [iCal]

With a growing interest in interpretability, there is an increasing need to characterize what exactly we mean by it and how to sensibly compare the interpretability of different approaches. In this talk, I suggest that our current desire for "interpretability" is as vague as asking for "good predictions" -- a desire that. while entirely reasonable, must be formalized into concrete needs such as high average test performance (perhaps held-out likelihood is a good metric) or some kind of robust performance (perhaps sensitivity or specificity are more appropriate metrics). This objective of this talk is to start a conversation to do the same for interpretability: I will describe distinct, concrete objectives that all fall under the umbrella term of interpretability and how each objective suggests natural evaluation procedures. I will also describe highlight important open questions in the evaluation of interpretable models.

Joint work with Been Kim, and the product of discussions with countless collaborators and colleagues.

Fri 1:00 a.m. - 1:30 a.m. [iCal]

Title: An unexpected unity among methods for interpreting model predictions Scott Lundberg and Su-In Lee

Title: Feature Importance Measure for Non-linear Learning Algorithms Marina M.-C. Vidovic, Nico Görnitz, Klaus-Robert Müller, and Marius Kloft

Fri 2:00 a.m. - 2:30 a.m. [iCal]

In machine learning often a tradeoff must be made between accuracy and intelligibility: the most accurate models usually are not very intelligible (e.g., deep neural nets, boosted trees, and random forests), and the most intelligible models usually are less accurate (e.g., linear/logistic regression). This tradeoff often limits the accuracy of models that can be applied in mission-critical applications such as healthcare where being able to understand, validate, edit, and ultimately trust a learned model is important. We have developed a learning method based on generalized additive models (GAMs) that is often as accurate as full complexity models, but remains as intelligible as linear/logistic regression models. In the talk I’ll present two case studies where these high-performance generalized additive models (GA2Ms) are applied to healthcare problems yielding intelligible models with state-of-the-art accuracy. In the pneumonia risk prediction case study, the intelligible model uncovers surprising patterns in the data that previously prevented complex learned models from being deployed, but because it is intelligible and modular allows these patterns to easily be recognized and removed. In the 30-day hospital readmission case study, we show that the same methods scale to large datasets containing hundreds of thousands of patients and thousands of attributes while remaining intelligible and providing accuracy comparable to the best (unintelligible) machine learning methods.

Fri 5:30 a.m. - 6:00 a.m. [iCal]
Maya Gupta (Invited Talk)
Fri 5:30 a.m. - 6:00 a.m. [iCal]

​What prior knowledge do humans have about machine learning problems that we can take advantage of as regularizers? One common intuition is that certain inputs should have a positive (only) effect on the output, for example, the price of a house should only increase as its size goes up, if all else is the same. Incorporating such monotonic priors into our machine learning algorithms can dramatically increase their interpretability and debuggability. We'll discuss state-of-the-art algorithms to learn flexible monotonic functions, and share some stories about why monotonicity is such an important regularizer for practical problems where train and test samples are not IID, especially when learning from clicks.

Fri 6:30 a.m. - 7:00 a.m. [iCal]

In many problem settings, parameters are not merely sparse, but dependent in such a way that non-zero coefficients tend to cluster together. We refer to this form of dependency as region sparsity". Classical sparse regression methods, such as the lasso and automatic relevance determination (ARD), which models parameters as independent a priori, and therefore do not exploit such dependencies. Here we introduce a hierarchical model for smooth, region-sparse weight vectors and tensors in a linear regression setting. Our approach represents a hierarchical extension of the relevance determination framework, where we add a transformed Gaussian process to model the dependencies between the prior variances of regression weights. We combine this with a structured model of the prior variances of Fourier coefficients, which eliminates unnecessary high frequencies. The resulting prior encourages weights to be region-sparse in two different bases simultaneously. We develop Laplace approximation and Monte Carlo Markov Chain (MCMC) sampling to provide efficient inference for the posterior, and show substantial improvements over existing methods for both simulated and real fMRI datasets.

Fri 7:00 a.m. - 7:30 a.m. [iCal]
Poster session (Posters)
Fri 7:30 a.m. - 8:00 a.m. [iCal]

Machine learning is the product of both an algorithm and data. While machine learning research tends to focus on algorithmic advances, taking the data as given, machine learning practice is quite the opposite. Most of the influence practitioners have in using machine learning to build predictive models comes through interacting with data, including crafting the data used for training and examining results on new data to inform future iterations. In this talk, I will present tools and techniques we have been developing in the Machine Teaching Group at Microsoft Research to support the model building process. I will then discuss some of the open challenges and opportunities in improving the practice of machine learning.

Fri 8:00 a.m. - 9:00 a.m. [iCal]
Future Directions in Interpretable Machine Learning (Panel Discussion)

Author Information

Andrew Wilson (Cornell University)

I am a professor of machine learning at New York University.

Been Kim (Google Brain)
William Herlands (Carnegie Mellon University)

More from the Same Authors