The development of new classification and regression algorithms based on empirical risk minimization (ERM) over deep neural network hypothesis classes, coined Deep Learning, revolutionized the area of artificial intelligence, machine learning, and data analysis. More recently, these methods have been applied to the numerical solution of high dimensional PDEs with great success. In particular, recent simulations indicate that deep learning based algorithms are capable of overcoming the curse of dimensionality for the numerical solution of linear Kolmogorov PDEs. Kolmogorov PDEs have been widely used in models from engineering, finance, and the natural sciences. Nearly all approximation methods for Kolmogorov PDEs in the literature suffer under the curse of dimensionality. By contrast, in recent work by some of the authors it was shown that deep ReLU neural networks are capable of approximating solutions of Kolmogorov PDEs without incurring the curse of dimensionality. The present paper considerably strengthens these results by providing an analysis of the generalization error. In particular we show that for Kolmogorov PDEs with affine drift and diffusion coefficients and a given accuracy $\varepsilon>0$, ERM over deep neural network hypothesis classes of size scaling polynomially in the dimension $d$ and $\varepsilon^{-1}$ and with a number of training samples scaling polynomially in the dimension $d$ and $\varepsilon^{-1}$ approximates the solution of the Kolmogorov PDE to within accuracy $\varepsilon$ with high probability. We conclude that ERM over deep neural network hypothesis classes breaks the curse of dimensionality for the numerical solution of linear Kolmogorov PDEs with affine drift and diffusion coefficients. To the best of our knowledge this is the first rigorous mathematical result that proves the efficiency of deep learning methods for high dimensional problems.

Reward design, the problem of selecting an appropriate reward function for an AI system, is both critically important, as it encodes the task the system should perform, and challenging, as it requires reasoning about and understanding the agent's environment in detail. AI practitioners often iterate on the reward function for their systems in a trial-and-error process to get their desired behavior. Inverse reward design (IRD) is a preference inference method that infers a true reward function from an observed, possibly misspecified, proxy reward function. This allows the system to determine when it should trust its observed reward function and respond appropriately. This has been shown to avoid problems in reward design such as negative side-effects (omitting a seemingly irrelevant but important aspect of the task) and reward hacking (learning to exploit unanticipated loopholes). In this paper, we actively select the $\textit{set of proxy reward functions}$ available to the designer. This improves the quality of inference and simplifies the associated reward design problem. We present two types of queries: discrete queries, where the system designer chooses from a discrete set of reward functions, and feature queries, where the system queries the designer for weights on a small set of features. We evaluate this approach with experiments in a personal shopping assistant domain and a 2D navigation domain. We find that our approach leads to reduced regret at test time compared with vanilla IRD. Our results indicate that actively selecting the set of available reward functions is a promising direction to improve the efficiency and effectiveness of reward design.

We consider the problem of performing classification when only binary measurements of data are available. This situation may arise due to the need for extreme compression of data or in the interest of hardware efficiency [11, 17, 18, 1]. Despite this extremely coarse quantization of the data, one can still perform learning tasks, such as classification, with high accuracy. The authors of [23] recently proposed a classification method for binary data, which they show to be reasonably accurate and sufficiently simple to allow for theoretical analysis in certain settings. Additionally, the predicted class can be approximately understood as the class whose binarized training data most closely and frequently matches that of the test point. As this approach will be the foundation of the work presented here, we discuss it in detail in the next section. Interpretability of algorithms and the ability to explain predictions is of increasing importance as machine learning algorithms are applied to an expanding range of problems in areas such as medicine, criminal justice, and finance [3, 2, 24]. Decisions made based on algorithmic predictions can have profound repercussions for both participating individuals as well as society at large. A major drawback to complex models such as deep neural networks [20, 15, 8, 19] is that it is extremely difficult to explain how or why such algorithms arrive at a specific prediction, see e.g.

It has been experimentally observed in recent years that multi-layer artificial neural networks have a surprising ability to generalize, even when trained with far more parameters than observations. Is there a theoretical basis for this? The best available bounds on their metric entropy and associated complexity measures are essentially linear in the number of parameters, which is inadequate to explain this phenomenon. Here we examine the statistical risk (mean squared predictive error) of multi-layer networks with $\ell^1$-type controls on their parameters and with ramp activation functions (also called lower-rectified linear units). In this setting, the risk is shown to be upper bounded by $[(L^3 \log d)/n]^{1/2}$, where $d$ is the input dimension to each layer, $L$ is the number of layers, and $n$ is the sample size. In this way, the input dimension can be much larger than the sample size and the estimator can still be accurate, provided the target function has such $\ell^1$ controls and that the sample size is at least moderately large compared to $L^3\log d$. The heart of the analysis is the development of a sampling strategy that demonstrates the accuracy of a sparse covering of deep ramp networks. Lower bounds show that the identified risk is close to being optimal.

We explore the concept of co-design in the context of neural network verification. Specifically, we aim to train deep neural networks that not only are robust to adversarial perturbations but also whose robustness can be verified more easily. To this end, we identify two properties of network models - weight sparsity and so-called ReLU stability - that turn out to significantly impact the complexity of the corresponding verification task. We demonstrate that improving weight sparsity alone already enables us to turn computationally intractable verification problems into tractable ones. Then, improving ReLU stability leads to an additional 4-13x speedup in verification times. An important feature of our methodology is its "universality," in the sense that it can be used with a broad range of training procedures and verification approaches.

Performance metrics (error measures) are vital components of the evaluation frameworks in various fields. The intention of this study was to overview of a variety of performance metrics and approaches to their classification. The main goal of the study was to develop a typology that will help to improve our knowledge and understanding of metrics and facilitate their selection in machine learning regression, forecasting and prognostics. Based on the analysis of the structure of numerous performance metrics, we propose a framework of metrics which includes four (4) categories: primary metrics, extended metrics, composite metrics, and hybrid sets of metrics. The paper identified three (3) key components (dimensions) that determine the structure and properties of primary metrics: method of determining point distance, method of normalization, method of aggregation of point distances over a data set.

Risk management plays a central role in sequential decision-making problems, common in fields such as portfolio management [Lai et al., 2011], autonomous driving [Maurer et al., 2016], and healthcare [Parker, 2009]. A common risk-measure is the variance of the expected sum of rewards/costs and the mean-variance tradeoff function [Sobel, 1982; Mannor and Tsitsiklis, 2011] is one of the most widely used objective functions in risk-sensitive decision-making. Other risk-sensitive objectives have also been studied, for example, Borkar [2002] studied exponential utility functions, Tamar et al. [2012] experimented with the Sharpe Ratio measurement, Chow et al. [2018] studied value at risk (VaR) and mean-VaR optimization, Chow and Ghavamzadeh [2014], Tamar et al. [2015b], and Chow et al. [2018] investigated conditional value at risk (CVaR) and mean-CVaR optimization in a static setting, and Tamar et al. [2015a] investigated coherent risk for both linear and nonlinear system dynamics. Compared with other widely used performance measurements, such as the Sharpe Ratio and CVaR, the mean-variance measurement has explicit interpretability and computational advantages [Markowitz et al., 2000; Li and Ng, 2000]. For example, the Sharpe Ratio tends to lead to solutions with less mean return [Tamar et al., 2012].

Learning strategies for imperfect information games from samples of interaction is a challenging problem. A common method for this setting, Monte Carlo Counterfactual Regret Minimization (MCCFR), can have slow long-term convergence rates due to high variance. In this paper, we introduce a variance reduction technique (VR-MCCFR) that applies to any sampling variant of MCCFR. Using this technique, per-iteration estimated values and updates are reformulated as a function of sampled values and state-action baselines, similar to their use in policy gradient reinforcement learning. The new formulation allows estimates to be bootstrapped from other estimates within the same episode, propagating the benefits of baselines along the sampled trajectory; the estimates remain unbiased even when bootstrapping from other estimates. Finally, we show that given a perfect baseline, the variance of the value estimates can be reduced to zero. Experimental evaluation shows that VR-MCCFR brings an order of magnitude speedup, while the empirical variance decreases by three orders of magnitude. The decreased variance allows for the first time CFR+ to be used with sampling, increasing the speedup to two orders of magnitude.

Multiple logic-based reconstruction of conceptual data modelling languages such as EER, UML Class Diagrams, and ORM exists. They mainly cover various fragments of the languages and none are formalised such that the logic applies simultaneously for all three modelling language families as unifying mechanism. This hampers interchangeability, interoperability, and tooling support. In addition, due to the lack of a systematic design process of the logic used for the formalisation, hidden choices permeate the formalisations that have rendered them incompatible. We aim to address these problems, first, by structuring the logic design process in a methodological way. We generalise and extend the DSL design process to apply to logic language design more generally and, in particular, by incorporating an ontological analysis of language features in the process. Second, availing of this extended process, of evidence gathered of language feature usage, and of computational complexity insights from Description Logics (DL), we specify logic profiles taking into account the ontological commitments embedded in the languages. The profiles characterise the minimum logic structure needed to handle the semantics of conceptual models, enabling the development of interoperability tools. There is no known DL language that matches exactly the features of those profiles and the common core is small (in the tractable $\mathcal{ALNI}$). Although hardly any inconsistencies can be derived with the profiles, it is promising for scalable runtime use of conceptual data models.

Although neural machine translation has achieved promising results, it suffers from slow translation speed. The direct consequence is that a trade-off has to be made between translation quality and speed, thus its performance can not come into full play. We apply cube pruning, a popular technique to speed up dynamic programming, into neural machine translation to speed up the translation. To construct the equivalence class, similar target hidden states are combined, leading to less RNN expansion operations on the target side and less \$\mathrm{softmax}\$ operations over the large target vocabulary. The experiments show that, at the same or even better translation quality, our method can translate faster compared with naive beam search by \$3.3\times\$ on GPUs and \$3.5\times\$ on CPUs.