We consider a novel application of inverse reinforcement learning which involves modeling, learning and predicting the commenting behavior of YouTube viewers. Each group of users is modeled as a rationally inattentive Bayesian agent. Our methodology integrates three key components. First, to identify distinct commenting patterns, we use deep embedded clustering to estimate framing information (essential extrinsic features) that clusters users into distinct groups. Second, we present an inverse reinforcement learning algorithm that uses Bayesian revealed preferences to test for rationality: does there exist a utility function that rationalizes the given data, and if yes, can it be used to predict future behavior? Finally, we impose behavioral economics constraints stemming from rational inattention to characterize the attention span of groups of users.The test imposes a R{\'e}nyi mutual information cost constraint which impacts how the agent can select attention strategies to maximize their expected utility. After a careful analysis of a massive YouTube dataset, our surprising result is that in most YouTube user groups, the commenting behavior is consistent with optimizing a Bayesian utility with rationally inattentive constraints. The paper also highlights how the rational inattention model can accurately predict future commenting behavior. The massive YouTube dataset and analysis used in this paper are available on GitHub and completely reproducible.

Recently, over-parameterized neural networks have been extensively analyzed in the literature. However, the previous studies cannot satisfactorily explain why fully trained neural networks are successful in practice. In this paper, we present a new theoretical framework for analyzing over-parameterized neural networks which we call neural feature repopulation. Our analysis can satisfactorily explain the empirical success of two level neural networks that are trained by standard learning algorithms. Our key theoretical result is that in the limit of infinite number of hidden neurons, over-parameterized two-level neural networks trained via the standard (noisy) gradient descent learns a well-defined feature distribution (population), and the limiting feature distribution is nearly optimal for the underlying learning task under certain conditions. Empirical studies confirm that predictions of our theory are consistent with the results observed in real practice.

We propose a method for causal inference using satellite image time series, in order to determine the treatment effects of interventions which impact climate change, such as deforestation. Simply put, the aim is to quantify the 'before versus after' effect of climate related human driven interventions, such as urbanization; as well as natural disasters, such as hurricanes and forest fires. As a concrete example, we focus on quantifying forest tree cover change/ deforestation due to human led causes. The proposed method involves the following steps. First, we uae computer vision and machine learning/deep learning techniques to detect and quantify forest tree coverage levels over time, at every time epoch. We then look at this time series to identify changepoints. Next, we estimate the expected (forest tree cover) values using a Bayesian structural causal model and projecting/forecasting the counterfactual. This is compared to the values actually observed post intervention, and the difference in the two values gives us the effect of the intervention (as compared to the non intervention scenario, i.e. what would have possibly happened without the intervention). As a specific use case, we analyze deforestation levels before and after the hyperinflation event (intervention) in Brazil (which ended in 1993-94), for the Amazon rainforest region, around Rondonia, Brazil. For this deforestation use case, using our causal inference framework can help causally attribute change/reduction in forest tree cover and increasing deforestation rates due to human activities at various points in time.

Multimodal fusion frameworks for Human Action Recognition (HAR) using depth and inertial sensor data have been proposed over the years. In most of the existing works, fusion is performed at a single level (feature level or decision level), missing the opportunity to fuse rich mid-level features necessary for better classification. To address this shortcoming, in this paper, we propose three novel deep multilevel multimodal fusion frameworks to capitalize on different fusion strategies at various stages and to leverage the superiority of multilevel fusion. At input, we transform the depth data into depth images called sequential front view images (SFIs) and inertial sensor data into signal images. Each input modality, depth and inertial, is further made multimodal by taking convolution with the Prewitt filter. Creating "modality within modality" enables further complementary and discriminative feature extraction through Convolutional Neural Networks (CNNs). CNNs are trained on input images of each modality to learn low-level, high-level and complex features. Learned features are extracted and fused at different stages of the proposed frameworks to combine discriminative and complementary information. These highly informative features are served as input to a multi-class Support Vector Machine (SVM). We evaluate the proposed frameworks on three publicly available multimodal HAR datasets, namely, UTD Multimodal Human Action Dataset (MHAD), Berkeley MHAD, and UTD-MHAD Kinect V2. Experimental results show the supremacy of the proposed fusion frameworks over existing methods.

In graph neural networks (GNNs), pooling operators compute local summaries of input graphs to capture their global properties; in turn, they are fundamental operators for building deep GNNs that learn effective, hierarchical representations. In this work, we propose the Node Decimation Pooling (NDP), a pooling operator for GNNs that generates coarsened versions of a graph by leveraging on its topology only. During training, the GNN learns new representations for the vertices and fits them to a pyramid of coarsened graphs, which is computed in a pre-processing step. As theoretical contributions, we first demonstrate the equivalence between the MAXCUT partition and the node decimation procedure on which NDP is based. Then, we propose a procedure to sparsify the coarsened graphs for reducing the computational complexity in the GNN; we also demonstrate that it is possible to drop many edges without significantly altering the graph spectra of coarsened graphs. Experimental results show that NDP grants a significantly lower computational cost once compared to state-of-the-art graph pooling operators, while reaching, at the same time, competitive accuracy performance on a variety of graph classification tasks.

Traditional event detection classifies a word or a phrase in a given sentence for a set of predefined event types. The limitation of such predefined set is that it prevents the adaptation of the event detection models to new event types. We study a novel formulation of event detection that describes types via several keywords to match the contexts in documents. This facilitates the operation of the models to new types. We introduce a novel feature-based attention mechanism for convolutional neural networks for event detection in the new formulation. Our extensive experiments demonstrate the benefits of the new formulation for new type extension for event detection as well as the proposed attention mechanism for this problem.

We propose a novel spectral convolutional neural network (CNN) model on graph structured data, namely Distributed Feedback-Looped Networks (DFNets). This model is incorporated with a robust class of spectral graph filters, called feedback-looped filters, to provide better localization on vertices, while still attaining fast convergence and linear memory requirements. Theoretically, feedback-looped filters can guarantee convergence w.r.t. a specified error bound, and be applied universally to any graph without knowing its structure. Furthermore, the propagation rule of this model can diversify features from the preceding layers to produce strong gradient flows. We have evaluated our model using two benchmark tasks: semi-supervised document classification on citation networks and semi-supervised entity classification on a knowledge graph. The experimental results show that our model considerably outperforms the state-of-the-art methods in both benchmark tasks over all datasets.

Financial fraud detection represents the challenge of finding anomalies in networks of financial transactions. In general, the anomaly detection (AD) is the problem of distinguishing between normal data samples with well defined patterns or signatures and those that do not conform to the expected profiles. The fraudulent behaviour in money laundering may manifest itself through unusual patterns in financial transaction networks. In such networks, nodes represents customers and the edges are transactions: a directed edge between two nodes illustrates that there is a money transfer in the respective direction, where the weight on the edge is the transferred amount. In this paper we present a survey on the fundamental anomaly detection techniques and then present briefly the relevant literature in connection with fraud detection context.

We begin by reiterating that common neural network activation functions have simple Bayesian origins. In this spirit, we go on to show that Bayes's theorem also implies a simple recurrence relation; this leads to a Bayesian recurrent unit with a prescribed feedback formulation. We show that introduction of a context indicator leads to a variable feedback that is similar to the forget mechanism in conventional recurrent units. A similar approach leads to a probabilistic input gate. The Bayesian formulation leads naturally to the two pass algorithm of the Kalman smoother or forward-backward algorithm, meaning that inference naturally depends upon future inputs as well as past ones. Experiments on speech recognition confirm that the resulting architecture can perform as well as a bidirectional recurrent network with the same number of parameters as a unidirectional one. Further, when configured explicitly bidirectionally, the architecture can exceed the performance of a conventional bidirectional recurrence.

Combining insights from machine learning and quantum Monte Carlo, the stochastic reconfiguration method with neural network Ansatz states is a promising new direction for high precision ground state estimation of quantum many body problems. At present, the method is heuristic, lacking a proper theoretical foundation. We initiate a thorough analysis of the learning landscape, and show that it reveals universal behavior reflecting a combination of the underlying physics and of the learning dynamics. In particular, the spectrum of the quantum Fisher matrix of complex restricted Boltzmann machine states can dramatically change across a phase transition. In contrast to the spectral properties of the quantum Fisher matrix, the actual weights of the network at convergence do not reveal much information about the system or the dynamics. Furthermore, we identify a new measure of correlation in the state by analyzing entanglement the eigenvectors. We show that, generically, the learning landscape modes with least entanglement have largest eigenvalue, suggesting that correlations are encoded in large flat valleys of the learning landscape, favoring stable representations of the ground state.