With the growing use of AI in highly consequential domains, the quantification and removal of bias with respect to protected attributes, such as gender, race, etc., is becoming increasingly important. While quantifying bias is essential, sometimes the needs of a business (e.g., hiring) may require the use of certain features that are critical in a way that any bias that can be explained by them might need to be exempted. E.g., a standardized test-score may be a critical feature that should be weighed strongly in hiring even if biased, whereas other features, such as zip code may be used only to the extent that they do not discriminate. In this work, we propose a novel information-theoretic decomposition of the total bias (in a counterfactual sense) into a non-exempt component that quantifies the part of the bias that cannot be accounted for by the critical features, and an exempt component which quantifies the remaining bias. This decomposition allows one to check if the bias arose purely due to the critical features (inspired from the business necessity defense of disparate impact law) and also enables selective removal of the non-exempt component if desired. We arrive at this decomposition through examples that lead to a set of desirable properties (axioms) that any measure of non-exempt bias should satisfy. We demonstrate that our proposed counterfactual measure satisfies all of them. Our quantification bridges ideas of causality, Simpson's paradox, and a body of work from information theory called Partial Information Decomposition. We also obtain an impossibility result showing that no observational measure of non-exempt bias can satisfy all of the desirable properties, which leads us to relax our goals and examine observational measures that satisfy only some of these properties. We then perform case studies to show how one can train models while reducing non-exempt bias.

A major part of the balance sheets of the largest US banks consists of credit card portfolios. Hence, managing the charge-off rates is a vital task for the profitability of the credit card industry. Different macroeconomic conditions affect individuals' behavior in paying down their debts. In this paper, we propose an expert system for loss forecasting in the credit card industry using macroeconomic indicators. We select the indicators based on a thorough review of the literature and experts' opinions covering all aspects of the economy, consumer, business, and government sectors. The state of the art machine learning models are used to develop the proposed expert system framework. We develop two versions of the forecasting expert system, which utilize different approaches to select between the lags added to each indicator. Among 19 macroeconomic indicators that were used as the input, six were used in the model with optimal lags, and seven indicators were selected by the model using all lags. The features that were selected by each of these models covered all three sectors of the economy. Using the charge-off data for the top 100 US banks ranked by assets from the first quarter of 1985 to the second quarter of 2019, we achieve mean squared error values of 1.15E-03 and 1.04E-03 using the model with optimal lags and the model with all lags, respectively. The proposed expert system gives a holistic view of the economy to the practitioners in the credit card industry and helps them to see the impact of different macroeconomic conditions on their future loss.

In this paper, we propose the novel model of combinatorial pure exploration with partial linear feedback (CPE-PL). In CPE-PL, given a combinatorial action space $\mathcal{X} \subseteq \{0,1\}^d$, in each round a learner chooses one action $x \in \mathcal{X}$ to play, obtains a random (possibly nonlinear) reward related to $x$ and an unknown latent vector $\theta \in \mathbb{R}^d$, and observes a partial linear feedback $M_{x} (\theta + \eta)$, where $\eta$ is a zero-mean noise vector and $M_x$ is a transformation matrix for $x$. The objective is to identify the optimal action with the maximum expected reward using as few rounds as possible. We also study the important subproblem of CPE-PL, i.e., combinatorial pure exploration with full-bandit feedback (CPE-BL), in which the learner observes full-bandit feedback (i.e. $M_x = x^{\top}$) and gains linear expected reward $x^{\top} \theta$ after each play. In this paper, we first propose a polynomial-time algorithmic framework for the general CPE-PL problem with novel sample complexity analysis. Then, we propose an adaptive algorithm dedicated to the subproblem CPE-BL with better sample complexity. Our work provides a novel polynomial-time solution to simultaneously address limited feedback, general reward function and combinatorial action space including matroids, matchings, and $s$-$t$ paths.

Classifiers that can be implemented on chip with minimal computational and memory resources are essential for edge computing in emerging applications such as medical and IoT devices. This paper introduces a machine learning model based on oblique decision trees to enable resource-efficient classification on a neural implant. By integrating model compression with probabilistic routing and implementing cost-aware learning, our proposed model could significantly reduce the memory and hardware cost compared to state-of-the-art models, while maintaining the classification accuracy. We trained the resource-efficient oblique tree with power-efficient regularization (ResOT-PE) on three neural classification tasks to evaluate the performance, memory, and hardware requirements. On seizure detection task, we were able to reduce the model size by 3.4X and the feature extraction cost by 14.6X compared to the ensemble of boosted trees, using the intracranial EEG from 10 epilepsy patients. In a second experiment, we tested the ResOT-PE model on tremor detection for Parkinson's disease, using the local field potentials from 12 patients implanted with a deep-brain stimulation (DBS) device. We achieved a comparable classification performance as the state-of-the-art boosted tree ensemble, while reducing the model size and feature extraction cost by 10.6X and 6.8X, respectively. We also tested on a 6-class finger movement detection task using ECoG recordings from 9 subjects, reducing the model size by 17.6X and feature computation cost by 5.1X. The proposed model can enable a low-power and memory-efficient implementation of classifiers for real-time neurological disease detection and motor decoding.

Despite impressive progress in the last decade, it still remains an open challenge to build models that generalize well across multiple tasks and datasets. One path to achieve this is to learn meaningful and compact representations, in which different semantic aspects of data are structurally disentangled. The focus of disentanglement approaches has been on separating independent factors of variation despite the fact that real-world observations are often not structured into meaningful independent causal variables to begin with. In this work we bridge the gap to real-world scenarios by analyzing the behavior of most prominent methods and disentanglement scores on correlated data in a large scale empirical study (including 3900 models). We show that systematically induced correlations in the dataset are being learned and reflected in the latent representations, while widely used disentanglement scores fall short of capturing these latent correlations. Finally, we demonstrate how to disentangle these latent correlations using weak supervision, even if we constrain this supervision to be causally plausible. Our results thus support the argument to learn independent mechanisms rather than independent factors of variations.

We study the problem of zero-order optimization of a strongly convex function. The goal is to find the minimizer of the function by a sequential exploration of its values, under measurement noise. We study the impact of higher order smoothness properties of the function on the optimization error and on the cumulative regret. To solve this problem we consider a randomized approximation of the projected gradient descent algorithm. The gradient is estimated by a randomized procedure involving two function evaluations and a smoothing kernel. We derive upper bounds for this algorithm both in the constrained and unconstrained settings and prove minimax lower bounds for any sequential search method. Our results imply that the zero-order algorithm is nearly optimal in terms of sample complexity and the problem parameters. Based on this algorithm, we also propose an estimator of the minimum value of the function achieving almost sharp oracle behavior. We compare our results with the state-of-the-art, highlighting a number of key improvements.

The scarcity of class-labeled data is a ubiquitous bottleneck in a wide range of machine learning problems. While abundant unlabeled data normally exist and provide a potential solution, it is extremely challenging to exploit them. In this paper, we address this problem by leveraging Positive-Unlabeled~(PU) classification and conditional generation with extra unlabeled data \emph{simultaneously}, both of which aim to make full use of agnostic unlabeled data to improve classification and generation performances. In particular, we present a novel training framework to jointly target both PU classification and conditional generation when exposing to extra data, especially out-of-distribution unlabeled data, by exploring the interplay between them: 1) enhancing the performance of PU classifiers with the assistance of a novel Conditional Generative Adversarial Network~(CGAN) that is robust to noisy labels, 2) leveraging extra data with predicted labels from a PU classifier to help the generation. Our key contribution is a Classifier-Noise-Invariant Conditional GAN~(CNI-CGAN) that can learn the clean data distribution from noisy labels predicted by a PU classifier. Theoretically, we proved the optimal condition of CNI-CGAN and experimentally, we conducted extensive evaluations on diverse datasets, verifying the simultaneous improvements on both classification and generation.

Learning with the \textit{instance-dependent} label noise is challenging, because it is hard to model such real-world noise. Note that there are psychological and physiological evidences showing that we humans perceive instances by decomposing them into parts. Annotators are therefore more likely to annotate instances based on the parts rather than the whole instances. Motivated by this human cognition, in this paper, we approximate the instance-dependent label noise by exploiting \textit{parts-dependent} label noise. Specifically, since instances can be approximately reconstructed by a combination of parts, we approximate the instance-dependent \textit{transition matrix} for an instance by a combination of the transition matrices for the parts of the instance. The transition matrices for parts can be learned by exploiting anchor points (i.e., data points that belong to a specific class almost surely). Empirical evaluations on synthetic and real-world datasets demonstrate our method is superior to the state-of-the-art approaches for learning from the instance-dependent label noise.

Trust Region Policy Optimization (TRPO) and Proximal Policy Optimization (PPO), as the widely employed policy based reinforcement learning (RL) methods, are prone to converge to a sub-optimal solution as they limit the policy representation to a particular parametric distribution class. To address this issue, we develop an innovative Optimistic Distributionally Robust Policy Optimization (ODRPO) algorithm, which effectively utilizes Optimistic Distributionally Robust Optimization (DRO) approach to solve the trust region constrained optimization problem without parameterizing the policies. Our algorithm improves TRPO and PPO with a higher sample efficiency and a better performance of the final policy while attaining the learning stability. Moreover, it achieves a globally optimal policy update that is not promised in the prevailing policy based RL algorithms. Experiments across tabular domains and robotic locomotion tasks demonstrate the effectiveness of our approach.

We study the problem of structuring a learned representation to significantly improve performance without supervision. Unlike most methods which focus on using side information like weak supervision or defining new regularization objectives, we focus on improving the learned representation by structuring the architecture of the model. We propose a self-attention based architecture to make the encoder explicitly associate parts of the representation with parts of the input observation. Meanwhile, our structural decoder architecture encourages a hierarchical structure in the latent space, akin to structural causal models, and learns a natural ordering of the latent mechanisms. We demonstrate how these models learn a representation which improves results in a variety of downstream tasks including generation, disentanglement, and transfer using several challenging and natural image datasets.