Rich Caruana

CLeaR Conference 2022 Conference Paper

Differentially Private Estimation of Heterogeneous Causal Effects

• Fengshi Niu
• Harsha Nori
• Brian Quistorff
• Rich Caruana
• Donald Ngwe
• Aadharsh Kannan

Estimating heterogeneous treatment effects in domains such as healthcare or social science often involves sensitive data where protecting privacy is important. We introduce a general meta-algorithm for estimating conditional average treatment effects (CATE) with differential privacy guarantees. Our meta-algorithm can work with simple, single-stage CATE estimators such as S-learner and more complex multi-stage estimators such as DR and R-learner. We perform a tight privacy analysis by taking advantage of sample splitting in our meta-algorithm and the parallel composition property of differential privacy. In this paper, we implement our approach using DP-EBMs as the base learner. DP-EBMs are interpretable, high-accuracy models with privacy guarantees, which allow us to directly observe the impact of DP noise on the learned causal model. Our experiments show that multi-stage CATE estimators incur larger accuracy loss than single-stage CATE or ATE estimators and that most of the accuracy loss from differential privacy is due to an increase in variance, not biased estimates of treatment effects.

ICLR Conference 2022 Conference Paper

NODE-GAM: Neural Generalized Additive Model for Interpretable Deep Learning

• Chun-Hao Chang
• Rich Caruana
• Anna Goldenberg

Deployment of machine learning models in real high-risk settings (e.g. healthcare) often depends not only on the model's accuracy but also on its fairness, robustness, and interpretability. Generalized Additive Models (GAMs) are a class of interpretable models with a long history of use in these high-risk domains, but they lack desirable features of deep learning such as differentiability and scalability. In this work, we propose a neural GAM (NODE-GAM) and neural GA$^2$M (NODE-GA$^2$M) that scale well and perform better than other GAMs on large datasets, while remaining interpretable compared to other ensemble and deep learning models. We demonstrate that our models find interesting patterns in the data. Lastly, we show that we are able to improve model accuracy via self-supervised pre-training, an improvement that is not possible for non-differentiable GAMs.

ICML Conference 2021 Conference Paper

Accuracy, Interpretability, and Differential Privacy via Explainable Boosting

• Harsha Nori
• Rich Caruana
• Zhiqi Bu
• Judy Hanwen Shen
• Janardhan Kulkarni

We show that adding differential privacy to Explainable Boosting Machines (EBMs), a recent method for training interpretable ML models, yields state-of-the-art accuracy while protecting privacy. Our experiments on multiple classification and regression datasets show that DP-EBM models suffer surprisingly little accuracy loss even with strong differential privacy guarantees. In addition to high accuracy, two other benefits of applying DP to EBMs are: a) trained models provide exact global and local interpretability, which is often important in settings where differential privacy is needed; and b) the models can be edited after training without loss of privacy to correct errors which DP noise may have introduced.

NeurIPS Conference 2021 Conference Paper

Neural Additive Models: Interpretable Machine Learning with Neural Nets

• Rishabh Agarwal
• Levi Melnick
• Nicholas Frosst
• Xuezhou Zhang
• Ben Lengerich
• Rich Caruana
• Geoffrey E. Hinton

Deep neural networks (DNNs) are powerful black-box predictors that have achieved impressive performance on a wide variety of tasks. However, their accuracy comes at the cost of intelligibility: it is usually unclear how they make their decisions. This hinders their applicability to high stakes decision-making domains such as healthcare. We propose Neural Additive Models (NAMs) which combine some of the expressivity of DNNs with the inherent intelligibility of generalized additive models. NAMs learn a linear combination of neural networks that each attend to a single input feature. These networks are trained jointly and can learn arbitrarily complex relationships between their input feature and the output. Our experiments on regression and classification datasets show that NAMs are more accurate than widely used intelligible models such as logistic regression and shallow decision trees. They perform similarly to existing state-of-the-art generalized additive models in accuracy, but are more flexible because they are based on neural nets instead of boosted trees. To demonstrate this, we show how NAMs can be used for multitask learning on synthetic data and on the COMPAS recidivism data due to their composability, and demonstrate that the differentiability of NAMs allows them to train more complex interpretable models for COVID-19.

NeurIPS Conference 2019 Conference Paper

Efficient Forward Architecture Search

• Hanzhang Hu
• John Langford
• Rich Caruana
• Saurajit Mukherjee
• Eric Horvitz
• Debadeepta Dey

We propose a neural architecture search (NAS) algorithm, Petridish, to iteratively add shortcut connections to existing network layers. The added shortcut connections effectively perform gradient boosting on the augmented layers. The proposed algorithm is motivated by the feature selection algorithm forward stage-wise linear regression, since we consider NAS as a generalization of feature selection for regression, where NAS selects shortcuts among layers instead of selecting features. In order to reduce the number of trials of possible connection combinations, we train jointly all possible connections at each stage of growth while leveraging feature selection techniques to choose a subset of them. We experimentally show this process to be an efficient forward architecture search algorithm that can find competitive models using few GPU days in both the search space of repeatable network modules (cell-search) and the space of general networks (macro-search). Petridish is particularly well-suited for warm-starting from existing models crucial for lifelong-learning scenarios.

AAAI Conference 2017 Conference Paper

Identifying Unknown Unknowns in the Open World: Representations and Policies for Guided Exploration

• Himabindu Lakkaraju
• Ece Kamar
• Rich Caruana
• Eric Horvitz

Predictive models deployed in the real world may assign incorrect labels to instances with high confidence. Such errors or unknown unknowns are rooted in model incompleteness, and typically arise because of the mismatch between training data and the cases encountered at test time. As the models are blind to such errors, input from an oracle is needed to identify these failures. In this paper, we formulate and address the problem of informed discovery of unknown unknowns of any given predictive model where unknown unknowns occur due to systematic biases in the training data. We propose a modelagnostic methodology which uses feedback from an oracle to both identify unknown unknowns and to intelligently guide the discovery. We employ a two-phase approach which first organizes the data into multiple partitions based on the feature similarity of instances and the confidence scores assigned by the predictive model, and then utilizes an explore-exploit strategy for discovering unknown unknowns across these partitions. We demonstrate the efficacy of our framework by varying the underlying causes of unknown unknowns across various applications. To the best of our knowledge, this paper presents the first algorithmic approach to the problem of discovering unknown unknowns of predictive models.

ICML Conference 2016 Conference Paper

Analysis of Deep Neural Networks with Extended Data Jacobian Matrix

• Shengjie Wang 0001
• Abdel-rahman Mohamed
• Rich Caruana
• Jeff A. Bilmes
• Matthai Philipose
• Matthew Richardson
• Krzysztof J. Geras
• Gregor Urban

Deep neural networks have achieved great successes on various machine learning tasks, however, there are many open fundamental questions to be answered. In this paper, we tackle the problem of quantifying the quality of learned wights of different networks with possibly different architectures, going beyond considering the final classification error as the only metric. We introduce \emphExtended Data Jacobian Matrix to help analyze properties of networks of various structures, finding that, the spectrum of the extended data jacobian matrix is a strong discriminating factor for networks of different structures and performance. Based on such observation, we propose a novel regularization method, which manages to improve the network performance comparably to dropout, which in turn verifies the observation.

AAAI Conference 2014 Conference Paper

Active Learning with Model Selection

• Alnur Ali
• Rich Caruana
• Ashish Kapoor

Most active learning methods avoid model selection by training models of one type (SVMs, boosted trees, etc.) using one pre-defined set of model hyperparameters. We propose an algorithm that actively samples data to simultaneously train a set of candidate models (different model types and/or different hyperparameters) and also select the best model from this set. The algorithm actively samples points for training that are most likely to improve the accuracy of the more promising candidate models, and also samples points for model selection— all samples count against the same labeling budget. This exposes a natural trade-off between the focused active sampling that is most effective for training models, and the unbiased sampling that is better for model selection. We empirically demonstrate on six test problems that this algorithm is nearly as effective as an active learning oracle that knows the optimal model in advance.

NeurIPS Conference 2014 Conference Paper

Do Deep Nets Really Need to be Deep?

• Jimmy Ba
• Rich Caruana

Currently, deep neural networks are the state of the art on problems such as speech recognition and computer vision. In this paper we empirically demonstrate that shallow feed-forward nets can learn the complex functions previously learned by deep nets and achieve accuracies previously only achievable with deep models. Moreover, in some cases the shallow nets can learn these deep functions using the same number of parameters as the original deep models. On the TIMIT phoneme recognition and CIFAR-10 image recognition tasks, shallow nets can be trained that perform similarly to complex, well-engineered, deeper convolutional models.

NeurIPS Conference 2013 Conference Paper

Using multiple samples to learn mixture models

• Jason Lee
• Ran Gilad-Bachrach
• Rich Caruana

In the mixture models problem it is assumed that there are $K$ distributions $\theta_{1}, \ldots, \theta_{K}$ and one gets to observe a sample from a mixture of these distributions with unknown coefficients. The goal is to associate instances with their generating distributions, or to identify the parameters of the hidden distributions. In this work we make the assumption that we have access to several samples drawn from the same $K$ underlying distributions, but with different mixing weights. As with topic modeling, having multiple samples is often a reasonable assumption. Instead of pooling the data into one sample, we prove that it is possible to use the differences between the samples to better recover the underlying structure. We present algorithms that recover the underlying structure under milder assumptions than the current state of art when either the dimensionality or the separation is high. The methods, when applied to topic modeling, allow generalization to words not present in the training data.

ICML Conference 2008 Conference Paper

An empirical evaluation of supervised learning in high dimensions

• Rich Caruana
• Nikolaos Karampatziakis
• Ainur Yessenalina

ICML Conference 2008 Conference Paper

Detecting statistical interactions with additive groves of trees

• Daria Sorokina
• Rich Caruana
• Mirek Riedewald
• Daniel Fink 0002

Discovering additive structure is an important step towards understanding a complex multi-dimensional function because it allows the function to be expressed as the sum of lower-dimensional components. When variables interact, however, their effects are not additive and must be modeled and interpreted simultaneously. We present a new approach for the problem of interaction detection. Our method is based on comparing the performance of unrestricted and restricted prediction models, where restricted models are prevented from modeling an interaction in question. We show that an additive model-based regression ensemble, Additive Groves, can be restricted appropriately for use with this framework, and thus has the right properties for accurately detecting variable interactions.

ICML Conference 2006 Conference Paper

An empirical comparison of supervised learning algorithms

• Rich Caruana
• Alexandru Niculescu-Mizil

A number of supervised learning methods have been introduced in the last decade. Unfortunately, the last comprehensive empirical evaluation of supervised learning was the Statlog Project in the early 90's. We present a large-scale empirical comparison between ten supervised learning methods: SVMs, neural nets, logistic regression, naive bayes, memory-based learning, random forests, decision trees, bagged trees, boosted trees, and boosted stumps. We also examine the effect that calibrating the models via Platt Scaling and Isotonic Regression has on their performance. An important aspect of our study is the use of a variety of performance criteria to evaluate the learning methods.

UAI Conference 2005 Conference Paper

Obtaining Calibrated Probabilities from Boosting

• Alexandru Niculescu-Mizil
• Rich Caruana

ICML Conference 2005 Conference Paper

Predicting good probabilities with supervised learning

• Alexandru Niculescu-Mizil
• Rich Caruana

We examine the relationship between the predictions made by different learning algorithms and true posterior probabilities. We show that maximum margin methods such as boosted trees and boosted stumps push probability mass away from 0 and 1 yielding a characteristic sigmoid shaped distortion in the predicted probabilities. Models such as Naive Bayes, which make unrealistic independence assumptions, push probabilities toward 0 and 1. Other models such as neural nets and bagged trees do not have these biases and predict well calibrated probabilities. We experiment with two ways of correcting the biased probabilities predicted by some learning methods: Platt Scaling and Isotonic Regression. We qualitatively examine what kinds of distortions these calibration methods are suitable for and quantitatively examine how much data they need to be effective. The empirical results show that after calibration boosted trees, random forests, and SVMs predict the best probabilities.

ICML Conference 2004 Conference Paper

Ensemble selection from libraries of models

• Rich Caruana
• Alexandru Niculescu-Mizil
• Geoff Crew
• Alex Ksikes

We present a method for constructing ensembles from libraries of thousands of models. Model libraries are generated using different learning algorithms and parameter settings. Forward stepwise selection is used to add to the ensemble the models that maximize its performance. Ensemble selection allows ensembles to be optimized to performance metric such as accuracy, cross entropy, mean precision, or ROC Area. Experiments with seven test problems and ten metrics demonstrate the benefit of ensemble selection.

JMLR Journal 2003 Journal Article

Benefitting from the Variables that Variable Selection Discards

• Rich Caruana
• Virginia R. de Sa

In supervised learning variable selection is used to find a subset of the available inputs that accurately predict the output. This paper shows that some of the variables that variable selection discards can beneficially be used as extra outputs for inductive transfer. Using discarded input variables as extra outputs forces the model to learn mappings from the variables that were selected as inputs to these extra outputs. Inductive transfer makes what is learned by these mappings available to the model that is being trained on the main output, often resulting in improved performance on that main output. We present three synthetic problems (two regression problems and one classification problem) where performance improves if some variables discarded by variable selection are used as extra outputs. We then apply variable selection to two real problems (DNA splice-junction and pneumonia risk prediction) and demonstrate the same effect: using some of the discarded input variables as extra outputs yields somewhat better performance on both of these problems than can be achieved by variable selection alone. This new approach enhances the benefit of variable selection by allowing the learner to benefit from variables that would otherwise have been discarded by variable selection, but without suffering the loss in performance that occurs when these variables are used as inputs.

NeurIPS Conference 2001 Conference Paper

(Not) Bounding the True Error

• John Langford
• Rich Caruana

We present a new approach to bounding the true error rate of a continuous valued classifier based upon PAC-Bayes bounds. The method first con- structs a distribution over classifiers by determining how sensitive each parameter in the model is to noise. The true error rate of the stochastic classifier found with the sensitivity analysis can then be tightly bounded using a PAC-Bayes bound. In this paper we demonstrate the method on artificial neural networks with results of a 
 order of magnitude im- provement vs. the best deterministic neural net bounds.

ICML Conference 2000 Conference Paper

FeatureBoost: A Meta-Learning Algorithm that Improves Model Robustness

• Joseph O'Sullivan
• John Langford 0001
• Rich Caruana
• Avrim Blum

NeurIPS Conference 2000 Conference Paper

Overfitting in Neural Nets: Backpropagation, Conjugate Gradient, and Early Stopping

• Rich Caruana
• Steve Lawrence
• C. Giles

The conventional wisdom is that backprop nets with excess hidden units generalize poorly. We show that nets with excess capacity generalize well when trained with backprop and early stopping. Experiments sug(cid: 173) gest two reasons for this: 1) Overfitting can vary significantly in different regions of the model. Excess capacity allows better fit to regions of high non-linearity, and backprop often avoids overfitting the regions of low non-linearity. 2) Regardless of size, nets learn task subcomponents in similar sequence. Big nets pass through stages similar to those learned by smaller nets. Early stopping can stop training the large net when it generalizes comparably to a smaller net. We also show that conjugate gradient can yield worse generalization because it overfits regions of low non-linearity when learning to fit regions of high non-linearity.

IROS Conference 1998 Conference Paper

Multitask pattern recognition for autonomous robots

• Rich Caruana
• Joseph O'Sullivan

We present a method called multitask pattern recognition (MTPR) that improves the accuracy and robustness of neural net based vision systems. The method trains neural nets on auxiliary recognition problems at the same time as the nets are trained on the main recognition task. The predictions the system makes for the auxiliary recognition tasks are not used, but the internal features learned for the auxiliary tasks improve performance on the main task. In addition, the auxiliary tasks allow us to focus the attention of learning towards image features it would otherwise ignore, thereby increasing the robustness of the learned models. We demonstrate MTPR on three problems (one synthetic, two real). On these problems MTPR improves performance 10%-30%. MTPR is applicable to many pattern recognition problems.

ICML Conference 1996 Conference Paper

Algorithms and Applications for Multitask Learning

• Rich Caruana

NeurIPS Conference 1996 Conference Paper

Promoting Poor Features to Supervisors: Some Inputs Work Better as Outputs

• Rich Caruana
• Virginia de

In supervised learning there is usually a clear distinction between inputs and outputs - inputs are what you will measure, outputs are what you will predict from those measurements. This paper shows that the distinction between inputs and outputs is not this simple. Some features are more useful as extra outputs than as inputs. By using a feature as an output we get more than just the case values but can. learn a mapping from the other inputs to that feature. For many features this mapping may be more useful than the feature value itself. We present two regression problems and one classification problem where performance improves if features that could have been used as inputs are used as extra outputs instead. This result is surprising since a feature used as an output is not used during testing.

ICML Conference 1995 Conference Paper

Removing the Genetics from the Standard Genetic Algorithm

• Shumeet Baluja
• Rich Caruana

NeurIPS Conference 1995 Conference Paper

Using the Future to "Sort Out" the Present: Rankprop and Multitask Learning for Medical Risk Evaluation

• Rich Caruana
• Shumeet Baluja
• Tom Mitchell

A patient visits the doctor; the doctor reviews the patient's history, asks questions, makes basic measurements (blood pressure, .. . ), and prescribes tests or treatment. The prescribed course of action is based on an assessment of patient risk-patients at higher risk are given more and faster attention. It is also sequential- it is too expensive to immediately order all tests which might later be of value. This paper presents two methods that together improve the accuracy of backprop nets on a pneumonia risk assessment problem by 10-50%. Rankprop improves on backpropagation with sum of squares error in ranking patients by risk. Multitask learning takes advantage of future lab tests available in the training set, but not available in practice when predictions must be made. Both methods are broadly applicable.

ICML Conference 1994 Conference Paper

Greedy Attribute Selection

• Rich Caruana
• Dayne Freitag

NeurIPS Conference 1994 Conference Paper

Learning Many Related Tasks at the Same Time with Backpropagation

• Rich Caruana

Hinton [6] proposed that generalization in artificial neural nets should improve if nets learn to represent the domain's underlying regularities. Abu-Mustafa's hints work [1] shows that the outputs of a backprop net can be used as inputs through which domain(cid: 173) specific information can be given to the net. We extend these ideas by showing that a backprop net learning many related tasks at the same time can use these tasks as inductive bias for each other and thus learn better. We identify five mechanisms by which multitask backprop improves generalization and give empirical evidence that multi task backprop generalizes better in real domains.

ICML Conference 1993 Conference Paper

Multitask Learning: A Knowledge-Based Source of Inductive Bias

• Rich Caruana

UAI Conference 1987 Conference Paper

The Automatic Training of Rule Bases that Use Numerical Uncertainty Representations

• Rich Caruana