The computation of convolution layers in deep neural networks typically rely on high performance routines that trade space for time by using additional memory (either for packing purposes or required as part of the algorithm) to improve performance. The problems with such an approach are two-fold. First, these routines incur additional memory overhead which reduces the overall size of the network that can fit on embedded devices with limited memory capacity. Second, these high performance routines were not optimized for performing convolution, which means that the performance obtained is usually less than conventionally expected. In this paper, we demonstrate that direct convolution, when implemented correctly, eliminates all memory overhead, and yields performance that is between 10% to 400% times better than existing high performance implementations of convolution layers on conventional and embedded CPU architectures. We also show that a high performance direct convolution exhibits better scaling performance, i.e. suffers less performance drop, when increasing the number of threads.

Forecasting the movements of stock prices is one the most challenging problems in financial markets analysis. In this paper, we use Machine Learning (ML) algorithms for the prediction of future price movements using limit order book data. Two different sets of features are combined and evaluated: handcrafted features based on the raw order book data and features extracted by ML algorithms, resulting in feature vectors with highly variant dimensionalities. Three classifiers are evaluated using combinations of these sets of features on two different evaluation setups and three prediction scenarios. Even though the large scale and high frequency nature of the limit order book poses several challenges, the scope of the conducted experiments and the significance of the experimental results indicate that Machine Learning highly befits this task carving the path towards future research in this field. Keywords: Machine Learning, limit order book, feature extraction, mid price forecasting 1. Introduction Forecasting of financial time series is a very challenging problem and has attracted scientific interest in the past few decades. Due to the inherently noisy and non-stationary nature of financial time series, statistical models are unsuitable for the task of modeling and forecasting such data. However, the lack of appropriate training and regularization algorithms for Neural Networks at the time, such as the dropout technique [6], rendered them susceptible to over fitting the training data. Support Vector Machines were deemed as better candidates for this task, as their solution implicitly involves the generalization error. The development of effective and efficient training algorithms for deeper architectures [7], in conjunction with the improved results such models presented, steered scientific interests towards Deep Learning techniques in many domains. Deep Learning methods are capable of modeling highly nonlinear, very complex data, making them suitable for application to financial data [8], as well as time series forecasting [9]. Furthermore, ML techniques which perform feature extraction may uncover robust features, better-suited to the specific task at hand. Autoencoders [10], are Neural Networks which learn new features extracted from the original input space, which can be used to enhance the performance of various tasks, such as classification or regression. Bag-of-Features (BoF) models comprise another feature extraction method that can be used to extract representations of objects described by multiple feature vectors, such as time-series [11].

While it has not yet been proven, empirical evidence suggests that model generalization is related to local properties of the optima which can be described via the Hessian. We connect model generalization with the local property of a solution under the PAC-Bayes paradigm. In particular, we prove that model generalization ability is related to the Hessian, the higher-order "smoothness" terms characterized by the Lipschitz constant of the Hessian, and the scales of the parameters. Guided by the proof, we propose a metric to score the generalization capability of the model, as well as an algorithm that optimizes the perturbed model accordingly.

We consider the task of multimodal music mood prediction based on the audio signal and the lyrics of a track. We reproduce the implementation of traditional feature engineering based approaches and propose a new model based on deep learning. We compare the performance of both approaches on a database containing 18,000 tracks with associated valence and arousal values and show that our approach outperforms classical models on the arousal detection task, and that both approaches perform equally on the valence prediction task. We also compare the a posteriori fusion with fusion of modalities optimized simultaneously with each unimodal model, and observe a significant improvement of valence prediction. We release part of our database for comparison purposes.

Real world experiments are expensive, and thus it is important to reach a target in minimum number of experiments. Experimental processes often involve control variables that changes over time. Such problems can be formulated as a functional optimisation problem. We develop a novel Bayesian optimisation framework for such functional optimisation of expensive black-box processes. We represent the control function using Bernstein polynomial basis and optimise in the coefficient space. We derive the theory and practice required to dynamically adjust the order of the polynomial degree, and show how prior information about shape can be integrated. We demonstrate the effectiveness of our approach for short polymer fibre design and optimising learning rate schedules for deep networks.

Learning visual feature representations for video analysis is a daunting task that requires a large amount of training samples and a proper generalization framework. Many of the current state of the art methods for video captioning and movie description rely on simple encoding mechanisms through recurrent neural networks to encode temporal visual information extracted from video data. In this paper, we introduce a novel multitask encoder-decoder framework for automatic semantic description and captioning of video sequences. In contrast to current approaches, our method relies on distinct decoders that train a visual encoder in a multitask fashion. Our system does not depend solely on multiple labels and allows for a lack of training data working even with datasets where only one single annotation is viable per video. Our method shows improved performance over current state of the art methods in several metrics on multi-caption and single-caption datasets. To the best of our knowledge, our method is the first method to use a multi-task approach for encoding video features. Our method demonstrates its robustness on the Large Scale Movie Description Challenge (LSMDC) 2017 where our method won the movie description task and its results were ranked among other competitors as the most helpful for the visually impaired.

Abstract--Convolutional Neural Networks (CNNs) are extremely computationally demanding, presenting a large barrier to their deployment on resource-constrained devices. Since such systems are where some of their most useful applications lie (e.g. In this paper we unify the two viewpoints in a Deep Learning Inference Stack and take an across-stack approach by implementing and evaluating the most common neural network compression techniques (weight pruning, channel pruning, and quantisation) and optimising their parallel execution with a range of programming approaches (OpenMP, OpenCL) and hardware architectures (CPU, GPU). We provide comprehensive Pareto curves to instruct tradeoffs under constraints of accuracy, execution time, and memory space. Recent years have yielded rapid advances in the field of deep learning, largely due to the unparalleled effectiveness of Convolutional Neural Networks (CNNs) on a variety of difficult problems [1]. These networks are designed to run on servers with negligible resource constraints, utilising powerful GPUs. As such, creative approaches are required to deploy them on hardware with limited resources in order to enable many useful applications (e.g. However, currently these optimisation approaches come with limited benchmarks and few comparisons. We outline a first step towards a more comprehensive understanding of the performance available under different constraints of inference accuracy, execution time, and memory space. Since [7] used CNNs to outperform more traditional statistical techniques on the ImageNet dataset [8] they have become a standard tool for image processing. With a growing ecosystem dedicated to training deep neural networks, the number of parameters that state-of-the-art networks demand has vastly increased; in 2012 the state-of-the-art, AlexNet, had 61M parameters spread over eight layers whereas the most recent ImageNet winner uses an ensemble of SENets [9], the largest of which has 115M parameters across 154 layers.

In this work, a new approach for Sun tracking systems is presented. Due to the current system limitations regarding costs and operational problems, a new approach based on low cost, computer vision open hardware and deep learning has been developed. The preliminary tests carried out successfully in Plataforma solar de Almeria (PSA), reveal the great potential and show the new approach as a good alternative to traditional systems. The proposed approach can provide key variables for the Sun tracking system control like cloud movements prediction, block and shadow detection, atmospheric attenuation or measures of concentrated solar radiation, which can improve the control strategies of the system and therefore the system performance.

Multiplicative noise, including dropout, is widely used to regularize deep neural networks (DNNs), and is shown to be effective in a wide range of architectures and tasks. From an information perspective, we consider injecting multiplicative noise into a DNN as training the network to solve the task with noisy information pathways, which leads to the observation that multiplicative noise tends to increase the correlation between features, so as to increase the signal-to-noise ratio of information pathways. However, high feature correlation is undesirable, as it increases redundancy in representations. In this work, we propose non-correlating multiplicative noise (NCMN), which exploits batch normalization to remove the correlation effect in a simple yet effective way. We show that NCMN significantly improves the performance of standard multiplicative noise on image classification tasks, providing a better alternative to dropout for batch-normalized networks. Additionally, we present a unified view of NCMN and shake-shake regularization, which explains the performance gain of the latter.

Starcraft II (SCII) is widely considered as the most challenging Real Time Strategy (RTS) game as of now, due to large observation space, huge (continuous and infinite) action space, partial observation, multi-player simultaneous game model, long time horizon decision, etc. To push the frontier of AI's capability, Deepmind and Blizzard jointly present the StarCraft II Learning Environment (SC2LE) --- a testbench for designing complex decision making systems. While SC2LE provides a few mini games such as \textit{MoveToBeacon}, \textit{CollectMineralShards}, and \textit{DefeatRoaches} where some AI agents achieve the professional player's level, it is still far away from achieving the professional level in a \emph{full} game. To initialize the research and investigation in the full game, we develop two AI agents --- the AI agent TStarBot1 is based on deep reinforcement learning over flat action structure, and the AI agent TStarBot2 is based on rule controller over hierarchical action structure. Both TStarBot1 and TStarBot2 are able to defeat the builtin AI agents from level 1 to level 10 in a full game (1v1 \textit{Zerg}-vs-\textit{Zerg} game on the AbyssalReef map), noting that level 8, level 9, and level 10 are cheating agents with full vision on the whole map, with resource harvest boosting, and with both, respectively \footnote{According to some informal discussions from the StarCraft II forum, level 10 builtin AI is estimated to be Platinum to Diamond~\cite{scii-forum}, which are equivalent to top 50\% - 30\% human players in the ranking system of Battle.net Leagues~\cite{liquid}. }.