Reinforcement learning with sparse rewards is still an open challenge. Classic methods rely on getting feedback via extrinsic rewards to train the agent, and in situations where this occurs very rarely the agent learns slowly or cannot learn at all. Similarly, if the agent receives also rewards that create suboptimal modes of the objective function, it will likely prematurely stop exploring. More recent methods add auxiliary intrinsic rewards to encourage exploration. However, auxiliary rewards lead to a non-stationary target for the Q-function. In this paper, we present a novel approach that (1) plans exploration actions far into the future by using a long-term visitation count, and (2) decouples exploration and exploitation by learning a separate function assessing the exploration value of the actions. Contrary to existing methods which use models of reward and dynamics, our approach is off-policy and model-free. We further propose new tabular environments for benchmarking exploration in reinforcement learning. Empirical results on classic and novel benchmarks show that the proposed approach outperforms existing methods in environments with sparse rewards, especially in the presence of rewards that create suboptimal modes of the objective function. Results also suggest that our approach scales gracefully with the size of the environment. Source code is available at https://github.com/sparisi/visit-value-explore

Smart grid is an alternative solution of the conventional power grid which harnesses the power of the information technology to save the energy and meet today's environment requirements. Due to the inherent vulnerabilities in the information technology, the smart grid is exposed to a wide variety of threats that could be translated into cyber-attacks. In this paper, we develop a deep learning-based intrusion detection system to defend against cyber-attacks in the advanced metering infrastructure network. The proposed machine learning approach is trained and tested extensively on an empirical industrial dataset which is composed of several attack categories including the scanning, buffer overflow, and denial of service attacks. Then, an experimental comparison in terms of detection accuracy is conducted to evaluate the performance of the proposed approach with Naive Bayes, Support Vector Machine, and Random Forest. The obtained results suggest that the proposed approaches produce optimal results comparing to the other algorithms. Finally, we propose a network architecture to deploy the proposed anomaly-based intrusion detection system across the Advanced Metering Infrastructure network. In addition, we propose a network security architecture composed of two types of Intrusion detection system types, Host and Network-based, deployed across the Advanced Metering Infrastructure network to inspect the traffic and detect the malicious one at all the levels.

We analyze the Gambler's problem, a simple reinforcement learning problem where the gambler has the chance to double or lose their bets until the target is reached. This is an early example introduced in the reinforcement learning textbook by Sutton and Barto (2018), where they mention an interesting pattern of the optimal value function with high-frequency components and repeating non-smooth points. It is however without further investigation. We provide the exact formula for the optimal value function for both the discrete and the continuous cases. Though simple as it might seem, the value function is pathological: fractal, self-similar, derivative taking either zero or infinity, not smooth on any interval, and not written as elementary functions. It is in fact one of the generalized Cantor functions, where it holds a complexity that has been uncharted thus far. Our analyses could lead insights into improving value function approximation, gradient-based algorithms, and Q-learning, in real applications and implementations.

Modern machine learning methods including deep learning have achieved great success in predictive accuracy for supervised learning tasks, but may still fall short in giving useful estimates of their predictive uncertainty. Quantifying uncertainty is especially critical in real-world settings, which often involve input distributions that are shifted from the training distribution due to a variety of factors including sample bias and non-stationarity. In such settings, well calibrated uncertainty estimates convey information about when a model's output should (or should not) be trusted. Many probabilistic deep learning methods, including Bayesian-and non-Bayesian methods, have been proposed in the literature for quantifying predictive uncertainty, but to our knowledge there has not previously been a rigorous large-scale empirical comparison of these methods under dataset shift. We present a large-scale benchmark of existing state-of-the-art methods on classification problems and investigate the effect of dataset shift on accuracy and calibration. We find that traditional post-hoc calibration does indeed fall short, as do several other previous methods. However, some methods that marginalize over models give surprisingly strong results across a broad spectrum of tasks.

Flow-based generative models parameterize probability distributions through an invertible transformation and can be trained by maximum likelihood. Invertible residual networks provide a flexible family of transformations where only Lipschitz conditions rather than strict architectural constraints are needed for enforcing invertibility. However, prior work trained invertible residual networks for density estimation by relying on biased log-density estimates whose bias increased with the network's expressiveness. We give a tractable unbiased estimate of the log density using a "Russian roulette" estimator, and reduce the memory required during training by using an alternative infinite series for the gradient. Furthermore, we improve invertible residual blocks by proposing the use of activation functions that avoid derivative saturation and generalizing the Lipschitz condition to induced mixed norms. The resulting approach, called Residual Flows, achieves state-of-theart performance on density estimation amongst flow-based models, and outperforms networks that use coupling blocks at joint generative and discriminative modeling.

Smart grid is an emerging and promising technology. It uses the power of information technologies to deliver intelligently the electrical power to customers, and it allows the integration of the green technology to meet the environmental requirements. Unfortunately, information technologies have its inherent vulnerabilities and weaknesses that expose the smart grid to a wide variety of security risks. The Intrusion detection system (IDS) plays an important role in securing smart grid networks and detecting malicious activity, yet it suffers from several limitations. Many research papers have been published to address these issues using several algorithms and techniques. Therefore, a detailed comparison between these algorithms is needed. This paper presents an overview of four data mining algorithms used by IDS in Smart Grid. An evaluation of performance of these algorithms is conducted based on several metrics including the probability of detection, probability of false alarm, probability of miss detection, efficiency, and processing time. Results show that Random Forest outperforms the other three algorithms in detecting attacks with higher probability of detection, lower probability of false alarm, lower probability of miss detection, and higher accuracy.

We propose an algorithm combining calibrated prediction and generalization bounds from learning theory to construct confidence sets for deep neural networks with PAC guarantees---i.e., the confidence set for a given input contains the true label with high probability. We demonstrate how our approach can be used to construct PAC confidence sets on ResNet for ImageNet, a visual object tracking model, and a dynamics model the half-cheetah reinforcement learning problem.

Abstract--In recent years, as intrusion attacks on IoT networks have grown exponentially, there is an immediate need for sophisticated intrusion detection systems (IDSs). A vast majority of current IDSs are data-driven, which means that one of the most important aspects of this area of research is the quality of the data acquired from IoT network traffic. Two of the most cited intrusion detection datasets are the KDDCup99 and the NSL-KDD. The main goal of our project was to conduct a robust comparison of both datasets by evaluating the performance of various Machine Learning (ML) classifiers trained on them with a larger set of classification metrics than previous researchers. From our research, we were able to conclude that the NSL-KDD dataset is of a higher quality than the KDDCup99 dataset as the classifiers trained on it were on average 20.18% less accurate. This is because the classifiers trained on the KDDCup99 dataset exhibited a bias towards the redundancies within it, allowing them to achieve higher accuracies. Index T erms --Intrusion Detection, Machine Learning, NSL-KDD, KDDCup99, Artificial Neural Network, Random Forest, Support Vector Machine, Naïve Bayes Classifier I. Introduction With the rapid development of the internet, intrusion attacks on IoT networks have been growing exponentially and are a highly pertinent threat in the modern era.

Sequence generation models are commonly refined with reinforcement learning over user-defined metrics. However, high gradient variance hinders the practical use of this method. To stabilize this method, we adapt to contextual generation of categorical sequences a policy gradient estimator, which evaluates a set of correlated Monte Carlo (MC) rollouts for variance control. Due to the correlation, the number of unique rollouts is random and adaptive to model uncertainty; those rollouts naturally become baselines for each other, and hence are combined to effectively reduce gradient variance. We also demonstrate the use of correlated MC rollouts for binary-tree softmax models, which reduce the high generation cost in large vocabulary scenarios by decomposing each categorical action into a sequence of binary actions. We evaluate our methods on both neural program synthesis and image captioning. The proposed methods yield lower gradient variance and consistent improvement over related baselines.

Normalizing flows transform a latent distribution through an invertible neural network for a flexible and pleasingly simple approach to generative modelling, while preserving an exact likelihood. We propose FlowGMM, an end-to-end approach to generative semi supervised learning with normalizing flows, using a latent Gaussian mixture model. FlowGMM is distinct in its simplicity, unified treatment of labelled and unlabelled data with an exact likelihood, interpretability, and broad applicability beyond image data. We show promising results on a wide range of applications, including AG-News and Yahoo Answers text data, tabular data, and semi-supervised image classification. We also show that FlowGMM can discover interpretable structure, provide real-time optimization-free feature visualizations, and specify well calibrated predictive distributions.