While all kinds of mixed data---from personal data, over panel and scientific data, to public and commercial data---are collected and stored, building probabilistic graphical models for these hybrid domains becomes more difficult. Users spend significant amounts of time in identifying the parametric form of the random variables (Gaussian, Poisson, Logit, etc.) involved and learning the mixed models. To make this difficult task easier, we propose the first trainable probabilistic deep architecture for hybrid domains that features tractable queries. It is based on Sum-Product Networks (SPNs) with piecewise polynomial leaf distributions together with novel nonparametric decomposition and conditioning steps using the Hirschfeld-Gebelein-Renyi Maximum Correlation Coefficient. This relieves the user from deciding a-priori the parametric form of the random variables but is still expressive enough to effectively approximate any distribution and permits efficient learning and inference.Our experiments show that the architecture, called Mixed SPNs, can indeed capture complex distributions across a wide range of hybrid domains.

Many applications infer the structure of a probabilistic graphical model from data to elucidate the relationships between variables. But how can we train graphical models on a massive data set? In this paper, we show how to construct coresets---compressed data sets which can be used as proxy for the original data and have provably bounded worst case error---for Gaussian dependency networks (DNs), i.e., cyclic directed graphical models over Gaussians, where the parents of each variable are its Markov blanket. Specifically, we prove that Gaussian DNs admit coresets of size independent of the size of the data set. Unfortunately, this does not extend to DNs over members of the exponential family in general. As we will prove, Poisson DNs do not admit small coresets. Despite this worst-case result, we will provide an argument why our coreset construction for DNs can still work well in practice on count data.To corroborate our theoretical results, we empirically evaluated the resulting Core DNs on real data sets. The results demonstrate significant gains over no or naive sub-sampling, even in the case of count data.

We develop Riemannian Stein Variational Gradient Descent (RSVGD), a Bayesian inference method that generalizes Stein Variational Gradient Descent (SVGD) to Riemann manifold. The benefits are two-folds: (i) for inference tasks in Euclidean spaces, RSVGD has the advantage over SVGD of utilizing information geometry, and (ii) for inference tasks on Riemann manifolds, RSVGD brings the unique advantages of SVGD to the Riemannian world. To appropriately transfer to Riemann manifolds, we conceive novel and non-trivial techniques for RSVGD, which are required by the intrinsically different characteristics of general Riemann manifolds from Euclidean spaces. We also discover Riemannian Stein's Identity and Riemannian Kernelized Stein Discrepancy. Experimental results show the advantages over SVGD of exploring distribution geometry and the advantages of particle-efficiency, iteration-effectiveness and approximation flexibility over other inference methods on Riemann manifolds.

Co-clustering computes clusters of data items and the related features concurrently, and it has been used in many applications such as community detection, product recommendation, computer vision, and pricing optimization. In this paper, we propose a new co-clustering method, called CoDiNMF, which improves the clustering quality and finds directional patterns among co-clusters by using multiple directed and undirected graphs. We design the objective function of co-clustering by using min-cut criterion combined with an additional term which controls the sum of net directional flow between different co-clusters. In addition, we show that a variant of Nonnegative Matrix Factorization (NMF) can solve the proposed objective function effectively. We run experiments on the US patents and BlogCatalog data sets whose ground truth have been known, and show that CoDiNMF improves clustering results compared to other co-clustering methods in terms of average F1 score, Rand index, and adjusted Rand index (ARI). Finally, we compare CoDiNMF and other co-clustering methods on the Wikipedia data set of philosophers, and we can find meaningful directional flow of influence among co-clusters of philosophers.

In this work we present classifier patching, an approach for adapting an existing black-box classification model to new data. Instead of creating a new model, patching infers regions in the instance space where the existing model is error-prone by training a classifier on the previously misclassified data. It then learns a specific model to determine the error regions, which allows to patch the old model’s predictions for them. Patching relies on a strong, albeit unchangeable, existing base classifier, and the idea that the true labels of seen instances will be available in batches at some point in time after the original classification. We experimentally evaluate our approach, and show that it meets the original design goals. Moreover, we compare our approach to existing methods from the domain of ensemble stream classification in both concept drift and transfer learning situations. Patching adapts quickly and achieves high classification accuracy, outperforming state-of-the-art competitors in either adaptation speed or accuracy in many scenarios.

Learning representations on Grassmann manifolds is popular in quite a few visual recognition tasks. In order to enable deep learning on Grassmann manifolds, this paper proposes a deep network architecture by generalizing the Euclidean network paradigm to Grassmann manifolds. In particular, we design full rank mapping layers to transform input Grassmannian data to more desirable ones, exploit re-orthonormalization layers to normalize the resulting matrices, study projection pooling layers to reduce the model complexity in the Grassmannian context, and devise projection mapping layers to respect Grassmannian geometry and meanwhile achieve Euclidean forms for regular output layers. To train the Grassmann networks, we exploit a stochastic gradient descent setting on manifolds of the connection weights, and study a matrix generalization of backpropagation to update the structured data. The evaluations on three visual recognition tasks show that our Grassmann networks have clear advantages over existing Grassmann learning methods, and achieve results comparable with state-of-the-art approaches.

In recent years, significant progress has been made in solving challenging problems across various domains using deep reinforcement learning (RL). Reproducing existing work and accurately judging the improvements offered by novel methods is vital to sustaining this progress. Unfortunately, reproducing results for state-of-the-art deep RL methods is seldom straightforward. In particular, non-determinism in standard benchmark environments, combined with variance intrinsic to the methods, can make reported results tough to interpret. Without significance metrics and tighter standardization of experimental reporting, it is difficult to determine whether improvements over the prior state-of-the-art are meaningful. In this paper, we investigate challenges posed by reproducibility, proper experimental techniques, and reporting procedures. We illustrate the variability in reported metrics and results when comparing against common baselines and suggest guidelines to make future results in deep RL more reproducible. We aim to spur discussion about how to ensure continued progress in the field by minimizing wasted effort stemming from results that are non-reproducible and easily misinterpreted.

A temporally abstract action, or an option, is specified by a policy and a termination condition: the policy guides the option behavior, and the termination condition roughly determines its length. Generally, learning with longer options (like learning with multi-step returns) is known to be more efficient. However, if the option set for the task is not ideal, and cannot express the primitive optimal policy well, shorter options offer more flexibility and can yield a better solution. Thus, the termination condition puts learning efficiency at odds with solution quality. We propose to resolve this dilemma by decoupling the behavior and target terminations, just like it is done with policies in off-policy learning. To this end, we give a new algorithm, Q(beta), that learns the solution with respect to any termination condition, regardless of how the options actually terminate. We derive Q(beta) by casting learning with options into a common framework with well-studied multi-step off policy learning. We validate our algorithm empirically, and show that it holds up to its motivating claims.

Regression over a stream of data is challenging due to unbounded data size and non-stationary distribution over time. Typically, a traditional supervised regression model over a data stream is trained on data instances occurring within a short time period by assuming a stationary distribution. This model is later used to predict value of response-variable in future instances. Over time, the model may degrade in performance due to changes in data distribution among incoming data instances. Updating the model for change adaptation requires true value for every recent data instances, which is scarce in practice. To overcome this issue, recent studies have employed techniques that sample fewer instances to be used for model retraining. Yet, this may introduce sampling bias that adversely affects the model performance. In this paper, we study the regression problem over data streams in a novel setting. We consider two independent, yet related, non-stationary data streams, which are referred to as the source and the target stream. The target stream continuously generates data instances whose value of response variable is unknown. The source stream, however, continuously generates data instances along with corresponding value for the response-variable, and has a biased data distribution with respect to the target stream. We refer to the problem of using a model trained on the biased source stream to predict the response-variable’s value in data instances occurring on the target stream as Multistream Regression. In this paper, we describe a framework for multistream regression that simultaneously overcomes distribution bias and detects change in data distribution represented by the two streams over time using a Gaussian kernel model. We analyze the theoretical properties of the proposed approach and empirically evaluate it on both real-world and synthetic data sets. Importantly, our results indicate superior performance by the framework compared to other baseline regression methods.

Many real-world problems, such as network packet routing and the coordination of autonomous vehicles, are naturally modelled as cooperative multi-agent systems. There is a great need for new reinforcement learning methods that can efficiently learn decentralised policies for such systems. To this end, we propose a new multi-agent actor-critic method called counterfactual multi-agent (COMA) policy gradients. COMA uses a centralised critic to estimate the Q-function and decentralised actors to optimise the agents' policies. In addition, to address the challenges of multi-agent credit assignment, it uses a counterfactual baseline that marginalises out a single agent's action, while keeping the other agents' actions fixed. COMA also uses a critic representation that allows the counterfactual baseline to be computed efficiently in a single forward pass. We evaluate COMA in the testbed of StarCraft unit micromanagement, using a decentralised variant with significant partial observability. COMA significantly improves average performance over other multi-agent actor-critic methods in this setting, and the best performing agents are competitive with state-of-the-art centralised controllers that get access to the full state.