Generalization is essential for deep learning. In contrast to previous works claiming that Deep Neural Networks (DNNs) have an implicit regularization implemented by the stochastic gradient descent, we demonstrate explicitly Bayesian regularizations in a specific category of DNNs, i.e., Convolutional Neural Networks (CNNs). First, we introduce a novel probabilistic representation for the hidden layers of CNNs and demonstrate that CNNs correspond to Bayesian networks with the serial connection. Furthermore, we show that the hidden layers close to the input formulate prior distributions, thus CNNs have explicitly Bayesian regularizations based on the Bayesian regularization theory. In addition, we clarify two recently observed empirical phenomena that are inconsistent with traditional theories of generalization. Finally, we validate the proposed theory on a synthetic dataset

Deductive formalisms have been strongly developed in recent years; among them, Answer Set Programming (ASP) gained some momentum, and has been lately fruitfully employed in many real-world scenarios. Nonetheless, in spite of a large number of success stories in relevant application areas, and even in industrial contexts, deductive reasoning cannot be considered the ultimate, comprehensive solution to AI; indeed, in several contexts, other approaches result to be more useful. Typical Bioinformatics tasks, for instance classification, are currently carried out mostly by Machine Learning (ML) based solutions. In this paper, we focus on the relatively new problem of analyzing the evolution of neurological disorders. In this context, ML approaches already demonstrated to be a viable solution for classification tasks; here, we show how ASP can play a relevant role in the brain evolution simulation task. In particular, we propose a general and extensible framework to support physicians and researchers at understanding the complex mechanisms underlying neurological disorders. The framework relies on a combined use of ML and ASP, and is general enough to be applied in several other application scenarios, which are outlined in the paper.

Since then, they have received continuously growing attention. ResNets have a recursive structure x k 1 x k R k( x k) where R k is a neural network and the copying of the input x k is called a skip connection. This structure can be seen as the explicit Euler discretisation of an associated ordinary differential equation (ODE) and this inspired intensive research. However, all of those works only consider the connection of ResNets to a relatively small class of ODEs. We show that by simultaneously increasing the number of skip connection as well as the expressivity of the networks R k the flow for an arbitrary right hand side f L 1 null I; C 0, 1 b (R d; R d)null can be approximated uniformly by deep ReLU ResNets on compact sets. Further, we derive estimates on the number of parameters needed to do this up to a prescribed accuracy under temporal regularity assumptions. We also give a self-contained introduction to the preliminaries regarding neural networks and differential equations. Here, we give an elementary proof for a quantitative universal approximation theorem for deep ReLU networks and see that weak ODEs with right hand side in L 1null I; C 0, 1 b (R d; R d)null are globally well posed. Finally, we discuss the possibility of using ResNets for diffeomorphic matching problems and propose some next steps in the theoretical foundation of this approach.

Abstract: We propose a joint simulation and real-world learning framework for mapping navigation instructions and raw first-person observations to continuous control. Our model estimates the need for environment exploration, predicts the likelihood of visiting environment positions during execution, and controls the agent to both explore and visit high-likelihood positions. We introduce Supervised Reinforcement Asynchronous Learning (SuReAL). Learning uses both simulation and real environments without requiring autonomous flight in the physical environment during training, and combines supervised learning for predicting positions to visit and reinforcement learning for continuous control. We evaluate our approach on a natural language instruction-following task with a physical quad-copter, and demonstrate effective execution and exploration behavior.

Collaborative filtering has witnessed significant improvem ents in recent years, largely due to models based on low-dimensional embeddings, like weighted matrix factorizati on (e.g., [26, 39]) and deep learning [23, 22, 33, 47, 62, 58, 20, 11], including autoencoders [58, 33]. Also neighborhoo d-based approaches are competitive in certain regimes (e.g., [1, 53, 54]), despite being simple heuristics based o n item-item (or user-user) similarity matrices (like cosin e similarity). In this paper, we outline that Markov Random Fi elds (MRF) are closely related to autoencoders as well as to neighborhood-based approaches. W e build on the enormo us progress made in learning MRFs, in particular in sparse inverse covariance estimation (e.g., [36, 59, 15, 2, 60, 44, 45, 63, 55, 24, 25, 52, 56, 51]). Much of the literature on sparse inverse covariance estimation focuses on the regi me where the number of data points n is much smaller than the number of variables m in the model ( n m).

To overcome the absence of training data for unseen classes, conventional zero-shot learning approaches mainly train their model on seen datapoints and leverage the semantic descriptions for both seen and unseen classes. Beyond exploiting relations between classes of seen and unseen, we present a deep generative model to provide the model with experience about both seen and unseen classes. Based on the variational auto-encoder with class-specific multi-modal prior, the proposed method learns the conditional distribution of seen and unseen classes. In order to circumvent the need for samples of unseen classes, we treat the non-existing data as missing examples. That is, our network aims to find optimal unseen datapoints and model parameters, by iteratively following the generating and learning strategy. Since we obtain the conditional generative model for both seen and unseen classes, classification as well as generation can be performed directly without any off-the-shell classifiers. In experimental results, we demonstrate that the proposed generating and learning strategy makes the model achieve the outperforming results compared to that trained only on the seen classes, and also to the several state-of-the-art methods.

We analysed a recently released dataset of scientific manuscripts that were either rejected or accepted from various conferences in artificial intelligence. We used a combination of semantic, lexical and psycholinguistic analyses of the full text of the manuscripts to compare them based on the outcome of the peer review process. We found that accepted manuscripts were written with words that are less frequent, that are acquired at an older age, and that are more abstract than rejected manuscripts. We also found that accepted manuscripts scored lower on two indicators of readability than rejected manuscripts, and that they also used more artificial intelligence jargon. An analysis of the references included in the manuscripts revealed that the subset of accepted submissions were more likely to cite the same publications. This finding was echoed by pairwise comparisons of the word content of the manuscripts (i.e. an indicator or semantic similarity), which was higher in the accepted manuscripts. Finally, we predicted the peer review outcome of manuscripts with their word content, with words related to machine learning and neural networks positively related with acceptance, whereas words related to logic, symbolic processing and knowledge-based systems negatively related with acceptance.

Automatic speech recognition (ASR) systems are of vital importance nowadays in commonplace tasks such as speech-to-text processing and language translation. This created the need of an ASR system that can operate in realistic crowded environments. Thus, speech enhancement is now considered as a fundamental building block in newly developed ASR systems. In this paper, a generative adversarial network (GAN) based framework is investigated for the task of speech enhancement of audio tracks. A new architecture based on CasNet generator and additional perceptual loss is incorporated to get realistically denoised speech phonetics. Finally, the proposed framework is shown to quantitatively outperform other GAN-based speech enhancement approaches.

Autoencoders, as a dimensionality reduction technique, have been recently applied to outlier detection. However, neural networks are known to be vulnerable to overfitting, and therefore have limited potential in the unsupervised outlier detection setting. Current approaches to ensemble-based autoencoders do not generate a sufficient level of diversity to avoid the overfitting issue. To overcome the aforementioned limitations we develop a Boosting-based Autoencoder Ensemble approach (in short, BAE). BAE is an unsupervised ensemble method that, similarly to the boosting approach, builds an adaptive cascade of autoencoders to achieve improved and robust results. BAE trains the autoencoder components sequentially by performing a weighted sampling of the data, aimed at reducing the amount of outliers used during training, and at injecting diversity in the ensemble. We perform extensive experiments and show that the proposed methodology outperforms state-of-the-art approaches under a variety of conditions.

It is well known that the problem of vanishing/exploding gradients is a challenge when training deep networks. In this paper, we describe another phenomenon, called vanishing nodes, that also increases the difficulty of training deep neural networks. As the depth of a neural network increases, the network's hidden nodes have more highly correlated behavior. This results in great similarities between these nodes. The redundancy of hidden nodes thus increases as the network becomes deeper. We call this problem vanishing nodes, and we propose the metric vanishing node indicator (VNI) for quantitatively measuring the degree of vanishing nodes. The VNI can be characterized by the network parameters, which is shown analytically to be proportional to the depth of the network and inversely proportional to the network width. The theoretical results show that the effective number of nodes vanishes to one when the VNI increases to one (its maximal value), and that vanishing/exploding gradients and vanishing nodes are two different challenges that increase the difficulty of training deep neural networks. The numerical results from the experiments suggest that the degree of vanishing nodes will become more evident during back-propagation training, and that when the VNI is equal to 1, the network cannot learn simple tasks (e.g. the XOR problem) even when the gradients are neither vanishing nor exploding. We refer to this kind of gradients as the walking dead gradients, which cannot help the network converge when having a relatively large enough scale. Finally, the experiments show that the likelihood of failed training increases as the depth of the network increases. The training will become much more difficult due to the lack of network representation capability.