United States

Blackwell approachability is a framework for reasoning about repeated games with vector-valued payoffs. We introduce predictive Blackwell approachability, where an estimate of the next payoff vector is given, and the decision maker tries to achieve better performance based on the accuracy of that estimator. In order to derive algorithms that achieve predictive Blackwell approachability, we start by showing a powerful connection between four well-known algorithms. Follow-the-regularized-leader (FTRL) and online mirror descent (OMD) are the most prevalent regret minimizers in online convex optimization. In spite of this prevalence, the regret matching (RM) and regret matching+ (RM+) algorithms have been preferred in the practice of solving large-scale games (as the local regret minimizers within the counterfactual regret minimization framework). We show that RM and RM+ are the algorithms that result from running FTRL and OMD, respectively, to select the halfspace to force at all times in the underlying Blackwell approachability game. By applying the predictive variants of FTRL or OMD to this connection, we obtain predictive Blackwell approachability algorithms, as well as predictive variants of RM and RM+. In experiments across 18 common zero-sum extensive-form benchmark games, we show that predictive RM+ coupled with counterfactual regret minimization converges vastly faster than the fastest prior algorithms (CFR+, DCFR, LCFR) across all games but two of the poker games, sometimes by two or more orders of magnitude.

AAAI 2021

Faster Game Solving via Predictive Blackwell Approachability: Connecting Regret Matching and Mirror Descent

technical paper

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

Neural network visualization techniques mark image locations by their relevancy to the network's classification. Existing methods are effective in highlighting the regions that affect the resulting classification the most. However, as we show, these methods are limited in their ability to identify the support for alternative classifications, an effect we name the saliency bias hypothesis. In this work, we integrate two lines of research: gradient-based methods and attribution-based methods, and develop an algorithm that provides per-class explainability.
The algorithm back-projects the per pixel local influence, in a manner that is guided by the local attributions, while correcting for salient features that would otherwise bias the explanation.
In an extensive battery of experiments, we demonstrate the ability of our methods to class-specific visualization, and not just the predicted label. Remarkably, the method obtains state of the art results in benchmarks that are commonly applied to gradient-based methods as well as in those that are employed mostly for evaluating attribution methods. Using a new unsupervised procedure, our method is also successful in demonstrating that self-supervised methods learn semantic information. Our code is available at: https://github.com/shirgur/AGFVisualization.

Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization

The estimation and inference of human predictive uncertainty have great potential to improve the sampling efficiency and prediction reliability of human-in-the-loop systems for smart healthcare, smart education, and human-computer interactions. Predictive uncertainty in humans is highly interpretable, but its measurement is poorly accessible. Contrarily, the predictive uncertainty of machine learning models, albeit with poor interpretability, is relatively easily accessible. Here, we demonstrate that the poor accessibility of human uncertainty can be resolved by exploiting simple and universally accessible deterministic neural networks. We propose a new model for human uncertainty inference, called proxy ensemble network (PEN). Simulations with a few benchmark datasets demonstrated that the model can efficiently learn human uncertainty from a small amount of data. To show its applicability in real-world problems, we performed behavioral experiments, in which 64 physicians classified medical images and reported their level of confidence. We showed that the PEN could predict both the uncertainty range and diagnoses given by subjects with high accuracy. Our results demonstrate the ability of machine learning in guiding human decision making; it can also help humans in learning more efficiently and accurately. To the best of our knowledge, this is the first study that explored the possibility of accessing human uncertainty via the lens of deterministic neural networks.

Human Uncertainty Inference via Deterministic Ensemble Neural Networks

Efficiently computing the weighted Jaccard similarity has become an active research topic in machine learning and theory. For sparse data, the standard technique is based on the consistent weighed sampling (CWS). For dense data, however, methods based on rejection sampling (RS) can be much more efficient. Nevertheless, existing RS methods are still slow for practical purposes. In this paper, we propose to improve RS by  a strategy, which we call  efficient rejection sampling (ERS), based on ``early stopping + densification''.  We analyze the statistical property of  ERS and provide experimental results to compare ERS with RS and other algorithms for hashing weighted Jaccard. The results demonstrate that ERS significantly improves the existing methods for estimating the weighted Jaccard similarity in relatively dense data.

Rejection Sampling for Weighted Jaccard Similarity Revisited

Asynchronous distributed algorithms are a popular way to reduce synchronization costs in large-scale optimization, and in particular for neural network training. However, for nonsmooth and nonconvex objectives, few convergence guarantees exist beyond cases where closed-form proximal operator solutions are available. 
As training most popular deep neural networks corresponds to optimizing nonsmooth and nonconvex objectives, there is a pressing need for such convergence guarantees. In this paper, we analyze for the first time the convergence of stochastic asynchronous optimization for this general class of objectives. In particular, we focus on stochastic subgradient methods allowing for block variable partitioning, where the shared model is asynchronously updated by concurrent processes. To this end, we use a probabilistic model which captures key features of real asynchronous scheduling between concurrent processes. Under this model, we establish convergence with probability one to an invariant set for stochastic subgradient methods with momentum. 
From a practical perspective, one issue with the family of algorithms that we consider is that they are not efficiently supported by machine learning frameworks, which mostly focus on distributed data-parallel strategies. 
To address this, we propose a new implementation strategy for  shared-memory based training of deep neural networks for a partitioned but shared model in single- and multi-GPU settings.  
Based on this implementation, we achieve on average1.2x speed-up in comparison to state-of-the-art training methods for popular image classification tasks, without compromising accuracy.

Asynchronous Optimization Methods for Efficient Training of Deep Neural Networks with Guarantees

Recently the concept of teaching has been introduced into machine learning, in which a teacher model is used to guide the training of a student model (which will be used in real tasks) through  data selection, loss function design, etc. Learning to reweight, which is a specific kind of teaching that reweights training data using a teacher model, receives much attention due to its simplicity and effectiveness. In existing learning to reweight works, the teacher model only utilizes shallow/surface information such as training iteration number and loss/accuracy of the student model from training/validation sets, but ignores the internal states of the student model, which limits the potential of learning to reweight. In this work, we propose an improved data reweighting algorithm, in which the student model provides its internal states to the teacher model, and the teacher model returns adaptive weights of training samples to enhance the training of the student model. The teacher model is jointly trained with the student model using meta gradients propagated from a validation set. Experiments on image classification with clean/noisy labels and neural machine translation empirically demonstrate that our algorithm makes significant improvement over previous methods.

Learning to Reweight with Deep Interactions

We study the problem faced by a decision maker who wants to locate a set of facilities on a real line and allocate agents/items to the facilities. The items have given locations on the line, and can only be assigned to one of their closest facilities. The facilities are controlled by managers, who have additive utility over the items. An optimal solution that maximizes the (utilitarian or egalitarian) social welfare of the facilities may present a very unbalanced allocation of the items to the facilities and hence be perceived as unfair. In this paper, we are interested in fair allocation among facility managers and consider the well-studied proportionality and envy-freeness fairness notions and their relaxations. We assess the availability, existence, approximability, and the quality (price of fairness) of fair solutions, where the quality measures the system efficiency loss under a fair allocation compared to the one that maximizes the social welfare. Further, we show that one can find a Pareto-optimal solution in polynomial time.

Facility’s Perspective to Fair Facility Location Problems

In this paper, we consider online continuous DR-submodular maximization with linear stochastic long-term constraints. Compared to the prior work on online submodular maximization, our setting introduces the extra complication of stochastic linear constraint functions that are i.i.d. generated at each round. In particular, at each time step a DR-submodular utility function and a constraint vector, i.i.d. generated from an unknown distribution, are revealed after committing to an action and we aim to maximize the overall utility while the expected cumulative resource consumption is below a fixed budget. Stochastic long-term constraints arise naturally in applications where there is a limited budget or resource available and resource consumption at each step is governed by stochastically time-varying environments. We propose the Online Lagrangian Frank-Wolfe (OLFW) algorithm to solve this class of online problems. We analyze the performance of the OLFW algorithm and we obtain sub-linear regret bounds as well as sub-linear cumulative constraint violation bounds, both in expectation and with high probability.

Online DR-Submodular Maximization: Minimizing Regret and Constraint Violation

Tensor decomposition is one of the most effective techniques for multi-criteria recommendations. However, it suffers from data sparsity when dealing with three-dimensional (3D) user-item-criterion ratings. To mitigate this issue, we consider effectively incorporating the side information and cross-domain knowledge in tensor decomposition. A deep transfer tensor decomposition (DTTD) method is proposed by integrating deep structure and Tucker decomposition, where an orthogonal constrained stacked denoising autoencoder (OC-SDAE) is proposed for alleviating the scale variation in learning effective latent representation, and the side information is incorporated as a compensation for tensor sparsity. Tucker decomposition generates private users and items' latent factors to connect with OC-SDAEs and creates a common core tensor to bridge different domains. A cross-domain alignment algorithm (CDAA) is proposed to solve the rotation issue between two core tensors in source and target domain. To the best of our knowledge, this is the first work in Tucker decomposition based recommendations to use deep structure to incorporate the side information and cross-domain knowledge. Experiments show that DTTD outperforms state-of-the-art related works.

Deep Transfer Tensor Decomposition with Orthogonal Constraint for Recommender Systems

The policy-based reinforcement learning (RL) can be considered as maximization of its objective. However, due to the inherent non-concavity of its objective, the policy gradient method to a first-order stationary point (FOSP) cannot guar- antee a maximal point. A FOSP can be a minimal or even a saddle point, which is undesirable for RL. It has be found that if all the saddle points are strict, all the second-order station- ary points (SOSP) are exactly equivalent to local maxima. Instead of FOSP, we consider SOSP as the convergence criteria to characterize the sample complexity of policy gradient. Our result shows that policy gradient converges to an (Îµ, âˆšÎµÏ‡)-SOSP with probability at least 1 âˆ’ O(Î´) after the total cost of O(Îµâˆ’9/2)sinificantly improves the state of the art cost O(Îµâˆ’9).Our analysis is based on the key idea that decomposes the parameter space Rp into three non-intersected regions: non-stationary point region, saddle point region, and local optimal region, then making a local improvement of the objective of RL in each region. This technique can be potentially generalized to extensive policy gradient methods. For the complete proof, please refer to https://arxiv.org/pdf/2012.01491.pdf.

Sample Complexity of Policy Gradient Finding Second-Order Stationary Points

Like humans, document summarization models can interpret a documentâ€™s contents in a number of ways. Unfortunately, the  neural models of today are largely black boxes that provide little explanation of how or why they generated a summary in the way they did. Therefore, to begin prying open the black box and to inject a level of control into the substance of the final summary, we developed a novel select-and-generate framework that focuses on explainability. By revealing the latent centrality and interactions between sentences, along with scores for novelty and relevance, users are given a window into the choices a model is making and an opportunity to guide those choices in a more desirable direction. A novel pair-wise matrix captures the sentence interactions, centrality and attribute scores, and a mask with tunable attribute thresholds allows the user to control which sentences are likely to be included in the extraction. A sentence-deployed attention mechanism in the abstractor ensures the final summary emphasizes the desired content. Additionally, the encoder is adaptable, supporting both Transformer- and BERT-based configurations. In a series of experiments assessed with ROUGE metrics and two human evaluations, ESCA outperformed eight state-of-the-art models on the CNN/DailyMail and NYT50 benchmark datasets.

Faster Game Solving via Predictive Blackwell Approachability: Connecting Regret Matching and Mirror Descent

Downloads

Next from AAAI 2021

Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Faster Game Solving via Predictive Blackwell Approachability: Connecting Regret Matching and Mirror Descent

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES

Downloads