Our paper proposes a generalization error bound for a general family of deep neural networks based on the spectral norm of weight matrices. Through introducing a novel characterization of the Lipschitz properties of neural network family, we achieve a tighter generalization error bound for ultra-deep neural networks, whose depth is much larger than the square root of its width. Besides the general deep neural networks, our results can be applied to derive new bounds for several popular architectures, including convolutional neural networks (CNNs), residual networks (ResNets), and hyperspherical networks (SphereNets). In the regime that the depth of these architectures is dominating, our bounds allow for the choice of much larger parameter spaces of weight matrices, inducing potentially stronger expressive ability.

Deep neural networks, and in particular recurrent networks, are promising candidates to control autonomous agents that interact in real-time with the physical world. However, this requires a seamless integration of temporal features into the network's architecture. For the training of and inference with recurrent neural networks, they are usually rolled out over time, and different rollouts exist. Conventionally, during inference the layers of a network are computed in a sequential manner resulting in sparse temporal integration of information and long response times. In this study, we present a theoretical framework to describe the set of all rollouts and demonstrate their differences in solving specific tasks. We prove that certain rollouts, also with only skip and no recurrent connections, enable earlier and more frequent responses, and show empirically that these early responses have better performance. The streaming rollout maximizes these properties and, in addition, enables a fully parallel execution of the network reducing the runtime on massively parallel devices. Additionally, we provide an open-source toolbox to design, train, evaluate, and online-interact with streaming rollouts.

Understanding cell identity is an important task in many biomedical areas. Expression patterns of specific marker genes have been used to characterize some limited cell types, but exclusive markers are not available for many cell types. A second approach is to use machine learning to discriminate cell types based on the whole gene expression profiles (GEPs). The accuracies of simple classification algorithms such as linear discriminators or support vector machines are limited due to the complexity of biological systems. We used deep neural networks to analyze 1040 GEPs from 16 different human tissues and cell types. After comparing different architectures, we identified a specific structure of deep autoencoders that can encode a GEP into a vector of 30 numeric values, which we call the cell identity code (CIC). The original GEP can be reproduced from the CIC with an accuracy comparable to technical replicates of the same experiment. Although we use an unsupervised approach to train the autoencoder, we show different values of the CIC are connected to different biological aspects of the cell, such as different pathways or biological processes. This network can use CIC to reproduce the GEP of the cell types it has never seen during the training. It also can resist some noise in the measurement of the GEP. Furthermore, we introduce classifier autoencoder, an architecture that can accurately identify cell type based on the GEP or the CIC.

In (Franceschi et al., 2018) we proposed a unified mathematical framework, grounded on bilevel programming, that encompasses gradient-based hyperparameter optimization and meta-learning. We formulated an approximate version of the problem where the inner objective is solved iteratively, and gave sufficient conditions ensuring convergence to the exact problem. In this work we show how to optimize learning rates, automatically weight the loss of single examples and learn hyper-representations with Far-HO, a software package based on the popular deep learning framework TensorFlow that allows to seamlessly tackle both HO and ML problems.

Video-game players generate huge amounts of data, as everything they do within a game is recorded. In particular, among all the stored actions and behaviors, there is information on the in-game purchases of virtual products. Such information is of critical importance in modern free-to-play titles, where gamers can select or buy a profusion of items during the game in order to progress and fully enjoy their experience. To try to maximize these kind of purchases, one can use a recommendation system so as to present players with items that might be interesting for them. Such systems can better achieve their goal by employing machine learning algorithms that are able to predict the rating of an item or product by a particular user. In this paper we evaluate and compare two of these algorithms, an ensemble-based model (extremely randomized trees) and a deep neural network, both of which are promising candidates for operational video-game recommender engines. Item recommenders can help developers improve the game. But, more importantly, it should be possible to integrate them into the game, so that users automatically get personalized recommendations while playing. The presented models are not only able to meet this challenge, providing accurate predictions of the items that a particular player will find attractive, but also sufficiently fast and robust to be used in operational settings.

We explore the use of deep learning and deep reinforcement learning for optimization problems in transportation. Many transportation system analysis tasks are formulated as an optimization problem - such as optimal control problems in intelligent transportation systems and long term urban planning. Often transportation models used to represent dynamics of a transportation system involve large data sets with complex input-output interactions and are difficult to use in the context of optimization. Use of deep learning metamodels can produce a lower dimensional representation of those relations and allow to implement optimization and reinforcement learning algorithms in an efficient manner. In particular, we develop deep learning models for calibrating transportation simulators and for reinforcement learning to solve the problem of optimal scheduling of travelers on the network.

Harshil Shah University College London Bowen Zheng University College London David Barber University College London & Alan Turing Institute Abstract We introduce the A ttentive Unsupervised T ext (W) riter (AUTR), which is a word level generative model for natural language. It uses a recurrent neural network with a dynamic attention and canvas memory mechanism to iteratively construct sentences. By viewing the state of the memory at intermediate stages and where the model is placing its attention, we gain insight into how it constructs sentences. We demonstrate that AUTR learns a meaningful latent representation for each sentence, and achieves competitive log-likelihood lower bounds whilst being computationally efficient. It is effective at generating and reconstructing sentences, as well as imputing missing words. 1 Introduction Latent variable models have recently enjoyed significant success when modelling images (Gregor et al. 2015; Rezende et al. 2016; Gulrajani et al. 2017), as well as sequential data such as handwriting and speech (Bayer and Osendorfer 2015; Chung et al. 2015). They specify a conditional distribution of observed data, given a set of hidden (latent) variables. The stochastic gradient varia-tional Bayes (SGVB) algorithm (Kingma and Welling 2014; Rezende, Mohamed, and Wierstra 2014) has made (approximate) maximum likelihood learning possible on a large scale in models where the true posterior distribution of the latent variables is not tractable.

The 2017 Fake News Challenge Stage 1 (FNC-1) shared task addressed a stance classification task as a crucial first step towards detecting fake news. To date, there is no in-depth analysis paper to critically discuss FNC-1's experimental setup, reproduce the results, and draw conclusions for next-generation stance classification methods. In this paper, we provide such an in-depth analysis for the three top-performing systems. We first find that FNC-1's proposed evaluation metric favors the majority class, which can be easily classified, and thus overestimates the true discriminative power of the methods. Therefore, we propose a new F1-based metric yielding a changed system ranking. Next, we compare the features and architectures used, which leads to a novel feature-rich stacked LSTM model that performs on par with the best systems, but is superior in predicting minority classes. To understand the methods' ability to generalize, we derive a new dataset and perform both in-domain and cross-domain experiments. Our qualitative and quantitative study helps interpreting the original FNC-1 scores and understand which features help improving performance and why. Our new dataset and all source code used during the reproduction study are publicly available for future research.

Neuromorphic computing, inspired by the brain, promises extreme efficiency for certain classes of learning tasks, such as classification and pattern recognition. The performance and power consumption of neuromorphic computing depends heavily on the choice of the neuron architecture. Digital neurons (Dig-N) are conventionally known to be accurate and efficient at high speed, while suffering from high leakage currents from a large number of transistors in a large design. On the other hand, analog/mixed-signal neurons are prone to noise, variability and mismatch, but can lead to extremely low-power designs. In this work, we will analyze, compare and contrast existing neuron architectures with a proposed mixed-signal neuron (MS-N) in terms of performance, power and noise, thereby demonstrating the applicability of the proposed mixed-signal neuron for achieving extreme energy-efficiency in neuromorphic computing. The proposed MS-N is implemented in 65 nm CMOS technology and exhibits > 100X better energy-efficiency across all frequencies over two traditional digital neurons synthesized in the same technology node. We also demonstrate that the inherent error-resiliency of a fully connected or even convolutional neural network (CNN) can handle the noise as well as the manufacturing non-idealities of the MS-N up to certain degrees. Notably, a system-level implementation on MNIST datasets exhibits a worst-case increase in classification error by 2.1% when the integrated noise power in the bandwidth is ~ 0.1 uV2, along with +-3{\sigma} amount of variation and mismatch introduced in the transistor parameters for the proposed neuron with 8-bit precision.

Generating humor and quotes are very challenging problems in the field of computational linguistics and are often tackled separately. In this paper, we present a controlled Long Short-Term Memory (LSTM) architecture which is trained with categorical data like jokes and quotes together by passing category as an input along with the sequence of words. The idea is that a single neural net will learn the structure of both jokes and quotes to generate them on demand according to input category. Importantly, we believe the neural net has more knowledge as it's trained on different datasets and hence will enable it to generate more creative jokes or quotes from the mixture of information. May the network generate a funny inspirational joke!