The deep layers of modern neural networks extract a rather rich set of features as an input propagates through the network. This paper sets out to harvest these rich intermediate representations for quantization with minimal accuracy loss while significantly reducing the memory footprint and compute intensity of the DNN. This paper utilizes knowledge distillation through teacher-student paradigm (Hinton et al., 2015) in a novel setting that exploits the feature extraction capability of DNNs for higher-accuracy quantization. As such, our algorithm logically divides a pretrained full-precision DNN to multiple sections, each of which exposes intermediate features to train a team of students independently in the quantized domain. This divide and conquer strategy, in fact, makes the training of each student section possible in isolation while all these independently trained sections are later stitched together to form the equivalent fully quantized network. Experiments on various DNNs (LeNet, ResNet-20, SVHN and VGG-11) show that, on average, this approach - called DCQ (Divide and Conquer Quantization) - achieves on average 9.7% accuracy improvement to a state-of-the-art quantized training technique, DoReFa (Zhou et al., 2016) for binary and ternary networks.

We develop a framework for interacting with uncertain environments in reinforcement learning (RL) by leveraging preferences in the form of utility functions. We claim that there is value in considering different risk measures during learning. In this framework, the preference for risk can be tuned by variation of the parameter $\beta$ and the resulting behavior can be risk-averse, risk-neutral or risk-taking depending on the parameter choice. We evaluate our framework for learning problems with model uncertainty. We measure and control for \emph{epistemic} risk using dynamic programming (DP) and policy gradient-based algorithms. The risk-averse behavior is then compared with the behavior of the optimal risk-neutral policy in environments with epistemic risk.

We present the first comprehensive study on automatic knowledge base construction for two prevalent commonsense knowledge graphs: ATOMIC (Sap et al., 2019) and ConceptNet (Speer et al., 2017). Contrary to many conventional KBs that store knowledge with canonical templates, commonsense KBs only store loosely structured open-text descriptions of knowledge. We posit that an important step toward automatic commonsense completion is the development of generative models of commonsense knowledge, and propose COMmonsEnse Transformers (COMET) that learn to generate rich and diverse commonsense descriptions in natural language. Despite the challenges of commonsense modeling, our investigation reveals promising results when implicit knowledge from deep pre-trained language models is transferred to generate explicit knowledge in commonsense knowledge graphs. Empirical results demonstrate that COMET is able to generate novel knowledge that humans rate as high quality, with up to 77.5% (ATOMIC) and 91.7% (ConceptNet) precision at top 1, which approaches human performance for these resources. Our findings suggest that using generative commonsense models for automatic commonsense KB completion could soon be a plausible alternative to extractive methods.

$Q$-learning with function approximation is one of the most popular methods in reinforcement learning. Though the idea of using function approximation was proposed at least $60$ years ago, even in the simplest setup, i.e, approximating $Q$-functions with linear functions, it is still an open problem how to design a provably efficient algorithm that learns a near-optimal policy. The key challenges are how to efficiently explore the state space and how to decide when to stop exploring in conjunction with the function approximation scheme. The current paper presents a provably efficient algorithm for $Q$-learning with linear function approximation. Under certain regularity assumptions, our algorithm, Difference Maximization $Q$-learning (DMQ), combined with linear function approximation, returns a near optimal policy using polynomial number of trajectories. Our algorithm introduces a new notion, the Distribution Shift Error Checking (DSEC) oracle. This oracle tests whether there exists a function in the function class that predicts well on a distribution $\mathcal{D}_1$, but predicts poorly on another distribution $\mathcal{D}_2$, where $\mathcal{D}_1$ and $\mathcal{D}_2$ are distributions over states induced by two different exploration policies. For the linear function class, this oracle is equivalent to solving a top eigenvalue problem. We believe our algorithmic insights, especially the DSEC oracle, are also useful in designing and analyzing reinforcement learning algorithms with general function approximation.

In developing countries like India agriculture plays an extremely important role in the lives of the population. In India, around 80\% of the population depend on agriculture or its by-products as the primary means for employment. Given large population dependency on agriculture, it becomes extremely important for the government to estimate market factors in advance and prepare for any deviation from those estimates. Commodity arrivals to market is an extremely important factor which is captured at district level throughout the country. Historical data and short-term prediction of important variables such as arrivals, prices, crop quality etc. for commodities are used by the government to take proactive steps and decide various policy measures. In this paper, we present a framework to work with short timeseries in conjunction with remote sensing data to predict future commodity arrivals. We deal with extremely high dimensional data which exceed the observation sizes by multiple orders of magnitude. We use cascaded layers of dimensionality reduction techniques combined with regularized regression models for prediction. We present results to predict arrivals to major markets and state wide prices for `Tur' (red gram) crop in Karnataka, India. Our model consistently beats popular ML techniques on many instances. Our model is scalable, time efficient and can be generalized to many other crops and regions. We draw multiple insights from the regression parameters, some of which are important aspects to consider when predicting more complex quantities such as prices in the future. We also combine the insights to generate important recommendations for different government organizations.

Building performance discrepancies between building design and operation are one of the causes that lead many new designs fail to achieve their goals and objectives. One of main factors contributing to the discrepancy is occupant behaviors. Occupants responding to a new design are influenced by several factors. Existing building performance models (BPMs) ignore or partially address those factors (called contextual factors) while developing BPMs. To potentially reduce the discrepancies and improve the prediction accuracy of BPMs, this paper proposes a computational framework for learning mixture models by using Generative Adversarial Networks (GANs) that appropriately combining existing BPMs with knowledge on occupant behaviors to contextual factors in new designs. Immersive virtual environments (IVEs) experiments are used to acquire data on such behaviors. Performance targets are used to guide appropriate combination of existing BPMs with knowledge on occupant behaviors. The resulting model obtained is called an augmented BPM. Two different experiments related to occupant lighting behaviors are shown as case study. The results reveal that augmented BPMs significantly outperformed existing BPMs with respect to achieving specified performance targets. The case study confirms the potential of the computational framework for improving prediction accuracy of BPMs during design.

Sequence-processing neural networks led to remarkable progress on many NLP tasks. As a consequence, there has been increasing interest in understanding to what extent they process language as humans do. We aim here to uncover which biases such models display with respect to "natural" word-order constraints. We train models to communicate about paths in a simple gridworld, using miniature languages that reflect or violate various natural language trends, such as the tendency to avoid redundancy or to minimize long-distance dependencies. We study how the controlled characteristics of our miniature languages affect individual learning and their stability across multiple network generations. The results draw a mixed picture. On the one hand, neural networks show a strong tendency to avoid long-distance dependencies. On the other hand, there is no clear preference for the efficient, non-redundant encoding of information that is widely attested in natural language. We thus suggest inoculating a notion of "effort" into neural networks, as a possible way to make their linguistic behavior more human-like.

Deep neural networks have achieved impressive performance in many applications but their large number of parameters lead to significant computational and storage overheads. Several recent works attempt to mitigate these overheads by designing compact networks using pruning of connections. However, we observe that most of the existing strategies to design compact networks fail to preserve network robustness against adversarial examples. In this work, we rigorously study the extension of network pruning strategies to preserve both benign accuracy and robustness of a network. Starting with a formal definition of the pruning procedure, including pre-training, weights pruning, and fine-tuning, we propose a new pruning method that can create compact networks while preserving both benign accuracy and robustness. Our method is based on two main insights: (1) we ensure that the training objectives of the pre-training and fine-tuning steps match the training objective of the desired robust model (e.g., adversarial robustness/verifiable robustness), and (2) we keep the pruning strategy agnostic to pre-training and fine-tuning objectives. We evaluate our method on four different networks on the CIFAR-10 dataset and measure benign accuracy, empirical robust accuracy, and verifiable robust accuracy. We demonstrate that our pruning method can preserve on average 93\% benign accuracy, 92.5\% empirical robust accuracy, and 85.0\% verifiable robust accuracy while compressing the tested network by 10$\times$.

We present in this paper experiments on Table Recognition in hand-written registry books. We first explain how the problem of row and column detection is modeled, and then compare two Machine Learning approaches (Conditional Random Field and Graph Convolutional Network) for detecting these table elements. Evaluation was conducted on death records provided by the Archive of the Diocese of Passau. Both methods show similar results, a 89 F1 score, a quality which allows for Information Extraction. Software and dataset are open source/data.

Smart Meters (SMs) are an important component of smart electrical grids, but they have also generated serious concerns about privacy data of consumers. In this paper, we present a general formulation of the privacy-preserving problem in SMs from an information-theoretic perspective. In order to capture the casual time series structure of the power measurements, we employ Directed Information (DI) as an adequate measure of privacy. On the other hand, to cope with a variety of potential applications of SMs data, we study different distortion measures along with the standard squared-error distortion. This formulation leads to a quite general training objective (or loss) which is optimized under a deep learning adversarial framework where two Recurrent Neural Networks (RNNs), referred to as the releaser and the attacker, are trained with opposite goals. An exhaustive empirical study is then performed to validate the proposed approach for different privacy problems in three actual data sets. Finally, we study the impact of the data mismatch problem, which occurs when the releaser and the attacker have different training data sets and show that privacy may not require a large level of distortion in real-world scenarios.