The gradient noise (GN) in the stochastic gradient descent (SGD) algorithm is often considered to be Gaussian in the large data regime by assuming that the \emph{classical} central limit theorem (CLT) kicks in. This assumption is often made for mathematical convenience, since it enables SGD to be analyzed as a stochastic differential equation (SDE) driven by a Brownian motion. We argue that the Gaussianity assumption might fail to hold in deep learning settings and hence render the Brownian motion-based analyses inappropriate. Inspired by non-Gaussian natural phenomena, we consider the GN in a more general context and invoke the \emph{generalized} CLT, which suggests that the GN converges to a \emph{heavy-tailed} $\alpha$-stable random vector, where \emph{tail-index} $\alpha$ determines the heavy-tailedness of the distribution. Accordingly, we propose to analyze SGD as a discretization of an SDE driven by a L\'{e}vy motion. Such SDEs can incur `jumps', which force the SDE and its discretization \emph{transition} from narrow minima to wider minima, as proven by existing metastability theory and the extensions that we proved recently. In this study, under the $\alpha$-stable GN assumption, we further establish an explicit connection between the convergence rate of SGD to a local minimum and the tail-index $\alpha$. To validate the $\alpha$-stable assumption, we conduct experiments on common deep learning scenarios and show that in all settings, the GN is highly non-Gaussian and admits heavy-tails. We investigate the tail behavior in varying network architectures and sizes, loss functions, and datasets. Our results open up a different perspective and shed more light on the belief that SGD prefers wide minima.

"Thinking in pictures," [1] i.e., spatial-temporal reasoning, effortless and instantaneous for humans, is believed to be a significant ability to perform logical induction and a crucial factor in the intellectual history of technology development. Modern Artificial Intelligence (AI), fueled by massive datasets, deeper models, and mighty computation, has come to a stage where (super-)human-level performances are observed in certain specific tasks. However, current AI's ability in "thinking in pictures" is still far lacking behind. In this work, we study how to improve machines' reasoning ability on one challenging task of this kind: Raven's Progressive Matrices (RPM). Specifically, we borrow the very idea of "contrast effects" from the field of psychology, cognition, and education to design and train a permutation-invariant model. Inspired by cognitive studies, we equip our model with a simple inference module that is jointly trained with the perception backbone. Combining all the elements, we propose the Contrastive Perceptual Inference network (CoPINet) and empirically demonstrate that CoPINet sets the new state-of-the-art for permutation-invariant models on two major datasets. We conclude that spatial-temporal reasoning depends on envisaging the possibilities consistent with the relations between objects and can be solved from pixel-level inputs.

Reinforcement Learning algorithms have recently been proposed to learn time-sequential control policies in the field of autonomous driving. Direct applications of Reinforcement Learning algorithms with discrete action space will yield unsatisfactory results at the operational level of driving where continuous control actions are actually required. In addition, the design of neural networks often fails to incorporate the domain knowledge of the targeting problem such as the classical control theories in our case. In this paper, we propose a hybrid model by combining Q-learning and classic PID (Proportion Integration Differentiation) controller for handling continuous vehicle control problems under dynamic driving environment. Particularly, instead of using a big neural network as Q-function approximation, we design a Quadratic Q-function over actions with multiple simple neural networks for finding optimal values within a continuous space. We also build an action network based on the domain knowledge of the control mechanism of a PID controller to guide the agent to explore optimal actions more efficiently.We test our proposed approach in simulation under two common but challenging driving situations, the lane change scenario and ramp merge scenario. Results show that the autonomous vehicle agent can successfully learn a smooth and efficient driving behavior in both situations.

In this paper, we propose DeepAlign, a novel approach to multi-perspective process anomaly correction, based on recurrent neural networks and bidirectional beam search. At the core of the DeepAlign algorithm are two recurrent neural networks trained to predict the next event. One is reading sequences of process executions from left to right, while the other is reading the sequences from right to left. By combining the predictive capabilities of both neural networks, we show that it is possible to calculate sequence alignments, which are used to detect and correct anomalies. DeepAlign utilizes the case-level and event-level attributes to closely model the decisions within a process. We evaluate the performance of our approach on an elaborate data corpus of 30 realistic synthetic event logs and compare it to three state-of-the-art conformance checking methods. DeepAlign produces better corrections than the rest of the field reaching an overall accuracy of 98.45% across all datasets, whereas the best comparable state-of-the-art method reaches 70.19%.

We present the first approach capable of learning domain-independent planning heuristics entirely from scratch. The heuristics we learn map the hypergraph representation of the delete-relaxation of the planning problem at hand, to a cost estimate that approximates that of the least-cost path from the current state to the goal through the hypergraph. We generalise Graph Networks to obtain a new framework for learning over hypergraphs, which we specialise to learn planning heuristics by training over state/value pairs obtained from optimal cost plans. Our experiments show that the resulting architecture, STRIPS-HGNs, is capable of learning heuristics that are competitive with existing delete-relaxation heuristics including LM-cut. We show that the heuristics we learn are able to generalise across different problems and domains, including to domains that were not seen during training.

The idea behind procedural content generation (PCG) in games is to create content automatically, using algorithms, instead of relying on user-designed content. While PCG approaches have traditionally focused on creating content for video games, they are now being applied to all kinds of virtual environments, thereby enabling training of machine learning systems that are significantly more general. For example, PCG's ability to generate never-ending streams of new levels has allowed DeepMind's Capture the Flag agent to reach beyond human-level-performance. Additionally, PCG-inspired methods such as domain randomization enabled OpenAI's robot arm to learn to manipulate objects with unprecedented dexterity. Level generation in 2D arcade games has also illuminated some shortcomings of standard deep RL methods, suggesting potential ways to train more general policies. This Review looks at key aspect of PCG approaches, including its ability to (1) enable new video games (such as No Man's Sky), (2) create open-ended learning environments, (3) combat overfitting in supervised and reinforcement learning tasks, and (4) create better benchmarks that could ultimately spur the development of better learning algorithms. We hope this article can introduce the broader machine learning community to PCG, which we believe will be a critical tool in creating a more general machine intelligence.

It is well known that deep generative models have a rich latent space, and that it is possible to smoothly manipulate their outputs by traversing this latent space. Recently, architectures have emerged that allow for more complex manipulations, such as making an image look as though it were from a different class, or painted in a certain style. These methods typically require large amounts of training in order to learn a single class of manipulations. We present Transflow Learning, a method for transforming a pre-trained generative model so that its outputs more closely resemble data that we provide afterwards. In contrast to previous methods, Transflow Learning does not require any training at all, and instead warps the probability distribution from which we sample latent vectors using Bayesian inference. Transflow Learning can be used to solve a wide variety of tasks, such as neural style transfer and few-shot classification.

Public road authorities and private mobility service providers need information derived from the current and predicted traffic states to act upon the daily urban system and its spatial and temporal dynamics. In this research, a real-time parking area state (occupancy, in- and outflux) prediction model (up to 60 minutes ahead) has been developed using publicly available historic and real time data sources. Based on a case study in a real-life scenario in the city of Arnhem, a Neural Network-based approach outperforms a Random Forest-based one on all assessed performance measures, although the differences are small. Both are outperforming a naive seasonal random walk model. Although the performance degrades with increasing prediction horizon, the model shows a performance gain of over 150% at a prediction horizon of 60 minutes compared with the naive model. Furthermore, it is shown that predicting the in- and outflux is a far more difficult task (i.e. performance gains of 30%) which needs more training data, not based exclusively on occupancy rate. However, the performance of predicting in- and outflux is less sensitive to the prediction horizon. In addition, it is shown that real-time information of current occupancy rate is the independent variable with the highest contribution to the performance, although time, traffic flow and weather variables also deliver a significant contribution. During real-time deployment, the model performs three times better than the naive model on average. As a result, it can provide valuable information for proactive traffic management as well as mobility service providers.

Given (a) poorly exposed image, DIF AR(c) produces an image with pleasing contrast and colour better matching the groundtruth (d) compared to the state-of-the-art DeepUPE model [42] (b). Abstract W e present a novel neural network architecture for the image signal processing (ISP) pipeline. In a camera system, the ISP is a critical component that forms a high quality RGB image from RA W camera sensor data. Typical ISP pipelines sequentially apply a complex set of traditional image processing modules, such as demosaicing, denoising, tone mapping, etc. W e introduce a new deep network that replaces all these modules, dubbed Deep Image Formation And Retouching (DIFAR) . DIF AR introduces a multi-scale context-aware pixel-level block for local de-noising/demosaicing operations and a retouching block for global refinement of image colour, luminance and saturation. DIF AR can also be trained for RGB to RGB image enhancement. DIF AR is parameter-efficient and outperforms recently proposed deep learning approaches in both objective and perceptual metrics, setting new state-of-the-art performance on multiple datasets including Samsung S7 [38] and MIT-Adobe 5k [6]. 1. Introduction Image quality is of fundamental importance in any imaging system, including DSLR and smartphone cameras. At the imaging sensor, RA W data is normally captured on a color filter array (such as the well-known Bayer pattern) where at each pixel, only a red, green, or blue color is available. This mosaiced RA W data suffers from noise, vignetting, lack of white balance, and many other defects and additionally has a high dynamic range. The camera's image signal processing (ISP) pipeline is responsible for forming a high quality RGB image with minimal noise, pleasing colors, sharp detail, and good contrast from the degraded RA W data. In most cases, the ISP is realised as a modular sequence of traditional image signal processing algorithms (Figure 2) each responsible for a single well-defined image operation (e.g.

Literature analysis facilitates researchers better understanding the development of science and technology. The conventional literature analysis focuses on the topics, authors, abstracts, keywords, references, etc., and rarely pays attention to the content of papers. In the field of machine learning, the involved methods (M) and datasets (D) are key information in papers. The extraction and mining of M and D are useful for discipline analysis and algorithm recommendation. In this paper, we propose a novel entity recognition model, called MDER, and constructe datasets from the papers of the PAKDD conferences (2009-2019). Some preliminary experiments are conducted to assess the extraction performance and the mining results are visualized.