Search

Autonomous aerial monitoring is an important task aimed at gathering information from areas that may not be easily accessible by humans. At the same time, this task often requires recognizing anomalies from a significant distance or not previously encountered in the past. In this paper, we propose a novel framework that leverages the advanced capabilities provided by Large Language Models (LLMs) to actively collect information and perform anomaly detection in novel scenes. To this end, we propose an LLM based model dialogue approach, in which two deep learning models engage in a dialogue to actively control a drone to increase perception and anomaly detection accuracy. We conduct our experiments in a high fidelity simulation environment where an LLM is provided with a predetermined set of natural language movement commands mapped into executable code functions. Additionally, we deploy a multimodal Visual Question Answering (VQA) model charged with the task of visual question answering and captioning. By engaging the two models in conversation, the LLM asks exploratory questions while simultaneously flying a drone into different parts of the scene, providing a novel way to implement active perception. By leveraging LLMs reasoning ability, we output an improved detailed description of the scene going beyond existing static perception approaches. In addition to information gathering, our approach is utilized for anomaly detection and our results demonstrate the proposed methods effectiveness in informing and alerting about potential hazards., Comment: Accepted to International Conference of Pattern Recognition (ICPR 2024)

This roadmap consolidates recent advances while exploring emerging applications, reflecting the remarkable diversity of hardware platforms, neuromorphic concepts, and implementation philosophies reported in the field. It emphasizes the critical role of cross-disciplinary collaboration in this rapidly evolving field.

Utilizing deep learning models to learn part-based representations holds significant potential for interpretable-by-design approaches, as these models incorporate latent causes obtained from feature representations through simple addition. However, training a part-based learning model presents challenges, particularly in enforcing non-negative constraints on the model's parameters, which can result in training difficulties such as instability and convergence issues. Moreover, applying such approaches in Deep Reinforcement Learning (RL) is even more demanding due to the inherent instabilities that impact many optimization methods. In this paper, we propose a non-negative training approach for actor models in RL, enabling the extraction of part-based representations that enhance interpretability while adhering to non-negative constraints. To this end, we employ a non-negative initialization technique, as well as a modified sign-preserving training method, which can ensure better gradient flow compared to existing approaches. We demonstrate the effectiveness of the proposed approach using the well-known Cartpole benchmark.

In this study a model pipeline is proposed that combines computer vision with control-theoretic methods and utilizes low cost sensors. The proposed work enables perception-aware motion control for a quadrotor UAV to detect and navigate to objects of interest such as wind turbines and electric towers. The distance to the object of interest was estimated utilizing RGB as the primary sensory input. For the needs of the study, the Microsoft AirSim simulator was used. As a first step, a YOLOv8 model was integrated providing the basic position setpoints towards the detection. From the YOLOv8 inference, a target yaw angle was derived. The subsequent algorithms, combining performant in computational terms computer vision methods and YOLOv8, actively drove the drone to measure the height of the detection. Based on the height, an estimate of the depth was retrieved. In addition to this step, a convolutional neural network was developed, namely ActvePerceptionNet aiming at active YOLOv8 inference. The latter was validated for wind turbines where the rotational motion of the propeller was found to affect object confidence in a near periodical fashion. The results of the simulation experiments conducted in this study showed efficient object height and distance estimation and effective localization., Comment: 10 pages

The detection of gravitational waves has revolutionized our understanding of the universe, offering unprecedented insights into its dynamics. A major goal of gravitational wave data analysis is to speed up the detection and parameter estimation process using machine learning techniques, in light of an anticipated surge in detected events that would render traditional methods impractical. Here, we present the first detections of new gravitational-wave candidate events in data from a network of interferometric detectors enabled by machine learning. We discuss several new enhancements of our ResNet-based deep learning code, AresGW, that increased its sensitivity, including a new hierarchical classification of triggers, based on different noise and frequency filters. The enhancements resulted in a significant reduction in the false alarm rate, allowing AresGW to surpass traditional pipelines in the number of detected events in its effective training range (single source masses between 7 and 50 solar masses and source chirp masses between 10 and 40 solar masses), when the new detections are included. We calculate the astrophysical significance of events detected with AresGW using a logarithmic ranking statistic and injections into O3 data. Furthermore, we present spectrograms, parameter estimation, and reconstruction in the time domain for our new candidate events and discuss the distribution of their properties. In addition, the AresGW code exhibited very good performance when tested across various two-detector setups and on observational data from the O1 and O2 observing periods. Our findings underscore the remarkable potential of AresGW as a fast and sensitive detection algorithm for gravitational-wave astronomy, paving the way for a larger number of future discoveries., Comment: 30 pages, 42 figures

Deep Learning (DL) has brought significant advances to robotics vision tasks. However, most existing DL methods have a major shortcoming, they rely on a static inference paradigm inherent in traditional computer vision pipelines. On the other hand, recent studies have found that active perception improves the perception abilities of various models by going beyond these static paradigms. Despite the significant potential of active perception, it poses several challenges, primarily involving significant changes in training pipelines for deep learning models. To overcome these limitations, in this work, we propose a generic supervised active perception pipeline for object detection that can be trained using existing off-the-shelf object detectors, while also leveraging advances in simulation environments. To this end, the proposed method employs an additional neural network architecture that estimates better viewpoints in cases where the object detector confidence is insufficient. The proposed method was evaluated on synthetic datasets, constructed within the Webots robotics simulator, showcasing its effectiveness in two object detection cases., Comment: 5 pages, 7 figures, 2023 IEEE Symposium Series on Computational Intelligence

Neuromorphic photonic accelerators are becoming increasingly popular, since they can significantly improve computation speed and energy efficiency, leading to femtojoule per MAC efficiency. However, deploying existing DL models on such platforms is not trivial, since a great range of photonic neural network architectures relies on incoherent setups and power addition operational schemes that cannot natively represent negative quantities. This results in additional hardware complexity that increases cost and reduces energy efficiency. To overcome this, we can train non-negative neural networks and potentially exploit the full range of incoherent neuromorphic photonic capabilities. However, existing approaches cannot achieve the same level of accuracy as their regular counterparts, due to training difficulties, as also recent evidence suggests. To this end, we introduce a methodology to obtain the non-negative isomorphic equivalents of regular neural networks that meet requirements of neuromorphic hardware, overcoming the aforementioned limitations. Furthermore, we also introduce a sign-preserving optimization approach that enables training of such isomorphic networks in a non-negative manner.

Financial portfolio management describes the task of distributing funds and conducting trading operations on a set of financial assets, such as stocks, index funds, foreign exchange or cryptocurrencies, aiming to maximize the profit while minimizing the loss incurred by said operations. Deep Learning (DL) methods have been consistently excelling at various tasks and automated financial trading is one of the most complex one of those. This paper aims to provide insight into various DL methods for financial trading, under both the supervised and reinforcement learning schemes. At the same time, taking into consideration sentiment information regarding the traded assets, we discuss and demonstrate their usefulness through corresponding research studies. Finally, we discuss commonly found problems in training such financial agents and equip the reader with the necessary knowledge to avoid these problems and apply the discussed methods in practice.

Even nowadays, where Deep Learning (DL) has achieved state-of-the-art performance in a wide range of research domains, accelerating training and building robust DL models remains a challenging task. To this end, generations of researchers have pursued to develop robust methods for training DL architectures that can be less sensitive to weight distributions, model architectures and loss landscapes. However, such methods are limited to adaptive learning rate optimizers, initialization schemes, and clipping gradients without investigating the fundamental rule of parameters update. Although multiplicative updates have contributed significantly to the early development of machine learning and hold strong theoretical claims, to best of our knowledge, this is the first work that investigate them in context of DL training acceleration and robustness. In this work, we propose an optimization framework that fits to a wide range of optimization algorithms and enables one to apply alternative update rules. To this end, we propose a novel multiplicative update rule and we extend their capabilities by combining it with a traditional additive update term, under a novel hybrid update method. We claim that the proposed framework accelerates training, while leading to more robust models in contrast to traditionally used additive update rule and we experimentally demonstrate their effectiveness in a wide range of task and optimization methods. Such tasks ranging from convex and non-convex optimization to difficult image classification benchmarks applying a wide range of traditionally used optimization methods and Deep Neural Network (DNN) architectures.

Energy time-series analysis describes the process of analyzing past energy observations and possibly external factors so as to predict the future. Different tasks are involved in the general field of energy time-series analysis and forecasting, with electric load demand forecasting, personalized energy consumption forecasting, as well as renewable energy generation forecasting being among the most common ones. Following the exceptional performance of Deep Learning (DL) in a broad area of vision tasks, DL models have successfully been utilized in time-series forecasting tasks. This paper aims to provide insight into various DL methods geared towards improving the performance in energy time-series forecasting tasks, with special emphasis in Greek Energy Market, and equip the reader with the necessary knowledge to apply these methods in practice., Comment: 13 papges, 4 figures

Uncertainty estimation is an important task for critical problems, such as robotics and autonomous driving, because it allows creating statistically better perception models and signaling the model's certainty in its predictions to the decision method or a human supervisor. In this paper, we propose a Variational Neural Network-based version of a Voxel Pseudo Image Tracking (VPIT) method for 3D Single Object Tracking. The Variational Feature Generation Network of the proposed Variational VPIT computes features for target and search regions and the corresponding uncertainties, which are later combined using an uncertainty-aware cross-correlation module in one of two ways: by computing similarity between the corresponding uncertainties and adding it to the regular cross-correlation values, or by penalizing the uncertain feature channels to increase influence of the certain features. In experiments, we show that both methods improve tracking performance, while penalization of uncertain features provides the best uncertainty quality., Comment: 5 pages, 2 figures, 1 table

Traditionally, gravitational waves are detected with techniques such as matched filtering or unmodeled searches based on wavelets. However, in the case of generic black hole binaries with non-aligned spins, if one wants to explore the whole parameter space, matched filtering can become impractical, which sets severe restrictions on the sensitivity and computational efficiency of gravitational-wave searches. Here, we use a novel combination of machine-learning algorithms and arrive at sensitive distances that surpass traditional techniques in a specific setting. Moreover, the computational cost is only a small fraction of the computational cost of matched filtering. The main ingredients are a 54-layer deep residual network (ResNet), a Deep Adaptive Input Normalization (DAIN), a dynamic dataset augmentation, and curriculum learning, based on an empirical relation for the signal-to-noise ratio. We compare the algorithm's sensitivity with two traditional algorithms on a dataset consisting of a large number of injected waveforms of non-aligned binary black hole mergers in real LIGO O3a noise samples. Our machine-learning algorithm can be used in upcoming rapid online searches of gravitational-wave events in a sizeable portion of the astrophysically interesting parameter space. We make our code, AResGW, and detailed results publicly available at https://github.com/vivinousi/gw-detection-deep-learning ., Comment: 10 pages, 11 figures, accepted for publication in PRD, code publicly available at https://github.com/vivinousi/gw-detection-deep-learning

In this paper we evaluate the impact of domain shift on human detection models trained on well known object detection datasets when deployed on data outside the distribution of the training set, as well as propose methods to alleviate such phenomena based on the available annotations from the target domain. Specifically, we introduce the OpenDR Humans in Field dataset, collected in the context of agricultural robotics applications, using the Robotti platform, allowing for quantitatively measuring the impact of domain shift in such applications. Furthermore, we examine the importance of manual annotation by evaluating three distinct scenarios concerning the training data: a) only negative samples, i.e., no depicted humans, b) only positive samples, i.e., only images which contain humans, and c) both negative and positive samples. Our results indicate that good performance can be achieved even when using only negative samples, if additional consideration is given to the training process. We also find that positive samples increase performance especially in terms of better localization. The dataset is publicly available for download at https://github.com/opendr-eu/datasets.

We present the results of the first Machine Learning Gravitational-Wave Search Mock Data Challenge (MLGWSC-1). For this challenge, participating groups had to identify gravitational-wave signals from binary black hole mergers of increasing complexity and duration embedded in progressively more realistic noise. The final of the 4 provided datasets contained real noise from the O3a observing run and signals up to a duration of 20 seconds with the inclusion of precession effects and higher order modes. We present the average sensitivity distance and runtime for the 6 entered algorithms derived from 1 month of test data unknown to the participants prior to submission. Of these, 4 are machine learning algorithms. We find that the best machine learning based algorithms are able to achieve up to 95% of the sensitive distance of matched-filtering based production analyses for simulated Gaussian noise at a false-alarm rate (FAR) of one per month. In contrast, for real noise, the leading machine learning search achieved 70%. For higher FARs the differences in sensitive distance shrink to the point where select machine learning submissions outperform traditional search algorithms at FARs $\geq 200$ per month on some datasets. Our results show that current machine learning search algorithms may already be sensitive enough in limited parameter regions to be useful for some production settings. To improve the state-of-the-art, machine learning algorithms need to reduce the false-alarm rates at which they are capable of detecting signals and extend their validity to regions of parameter space where modeled searches are computationally expensive to run. Based on our findings we compile a list of research areas that we believe are the most important to elevate machine learning searches to an invaluable tool in gravitational-wave signal detection., Comment: 25 pages, 6 figures, 4 tables, additional material available at https://github.com/gwastro/ml-mock-data-challenge-1

Deep learning methods have been employed in gravitational-wave astronomy to accelerate the construction of surrogate waveforms for the inspiral of spin-aligned black hole binaries, among other applications. We face the challenge of modeling the residual error of an artificial neural network that models the coefficients of the surrogate waveform expansion (especially those of the phase of the waveform) which we demonstrate has sufficient structure to be learnable by a second network. Adding this second network, we were able to reduce the maximum mismatch for waveforms in a validation set by 13.4 times. We also explored several other ideas for improving the accuracy of the surrogate model, such as the exploitation of similarities between waveforms, the augmentation of the training set, the dissection of the input space, using dedicated networks per output coefficient and output augmentation. In several cases, small improvements can be observed, but the most significant improvement still comes from the addition of a second network that models the residual error. Since the residual error for more general surrogate waveform models (when e.g., eccentricity is included) may also have a specific structure, one can expect our method to be applicable to cases where the gain in accuracy could lead to significant gains in computational time.

Knowledge Distillation has been established as a highly promising approach for training compact and faster models by transferring knowledge from heavyweight and powerful models. However, KD in its conventional version constitutes an enduring, computationally and memory demanding process. In this paper, Online Self-Acquired Knowledge Distillation (OSAKD) is proposed, aiming to improve the performance of any deep neural model in an online manner. We utilize k-nn non-parametric density estimation technique for estimating the unknown probability distributions of the data samples in the output feature space. This allows us for directly estimating the posterior class probabilities of the data samples, and we use them as soft labels that encode explicit information about the similarities of the data with the classes, negligibly affecting the computational cost. The experimental evaluation on four datasets validates the effectiveness of proposed method., Comment: Accepted at ICME 2021

In this paper, regularized lightweight deep convolutional neural network models, capable of effectively operating in real-time on devices with restricted computational power for high-resolution video input are proposed. Furthermore, a novel regularization method motivated by the Quadratic Mutual Information, in order to improve the generalization ability of the utilized models is proposed. Extensive experiments on various binary classification problems involved in autonomous systems are performed, indicating the effectiveness of the proposed models as well as of the proposed regularizer., Comment: Accepted at MLSP 2020

Semantic scene segmentation plays a critical role in a wide range of robotics applications, e.g., autonomous navigation. These applications are accompanied by specific computational restrictions, e.g., operation on low-power GPUs, at sufficient speed, and also for high-resolution input. Existing state-of-the-art segmentation models provide evaluation results under different setups and mainly considering high-power GPUs. In this paper, we investigate the behavior of the most successful semantic scene segmentation models, in terms of deployment (inference) speed, under various setups (GPUs, input sizes, etc.) in the context of robotics applications. The target of this work is to provide a comparative study of current state-of-the-art segmentation models so as to select the most compliant with the robotics applications requirements., Comment: Accepted at IISA 2021

Recently, artificial neural networks have been gaining momentum in the field of gravitational wave astronomy, for example in surrogate modelling of computationally expensive waveform models for binary black hole inspiral and merger. Surrogate modelling yields fast and accurate approximations of gravitational waves and neural networks have been used in the final step of interpolating the coefficients of the surrogate model for arbitrary waveforms outside the training sample. We investigate the existence of underlying structures in the empirical interpolation coefficients using autoencoders. We demonstrate that when the coefficient space is compressed to only two dimensions, a spiral structure appears, wherein the spiral angle is linearly related to the mass ratio. Based on this finding, we design a spiral module with learnable parameters, that is used as the first layer in a neural network, which learns to map the input space to the coefficients. The spiral module is evaluated on multiple neural network architectures and consistently achieves better speed-accuracy trade-off than baseline models. A thorough experimental study is conducted and the final result is a surrogate model which can evaluate millions of input parameters in a single forward pass in under 1ms on a desktop GPU, while the mismatch between the corresponding generated waveforms and the ground-truth waveforms is better than the compared baseline methods. We anticipate the existence of analogous underlying structures and corresponding computational gains also in the case of spinning black hole binaries.

In this paper, we propose 2D-Attention (2DA), a generic attention formulation for sequence data, which acts as a complementary computation block that can detect and focus on relevant sources of information for the given learning objective. The proposed attention module is incorporated into the recently proposed Neural Bag of Feature (NBoF) model to enhance its learning capacity. Since 2DA acts as a plug-in layer, injecting it into different computation stages of the NBoF model results in different 2DA-NBoF architectures, each of which possesses a unique interpretation. We conducted extensive experiments in financial forecasting, audio analysis as well as medical diagnosis problems to benchmark the proposed formulations in comparison with existing methods, including the widely used Gated Recurrent Units. Our empirical analysis shows that the proposed attention formulations can not only improve performances of NBoF models but also make them resilient to noisy data.

Knowledge Distillation (KD) methods are capable of transferring the knowledge encoded in a large and complex teacher into a smaller and faster student. Early methods were usually limited to transferring the knowledge only between the last layers of the networks, while latter approaches were capable of performing multi-layer KD, further increasing the accuracy of the student. However, despite their improved performance, these methods still suffer from several limitations that restrict both their efficiency and flexibility. First, existing KD methods typically ignore that neural networks undergo through different learning phases during the training process, which often requires different types of supervision for each one. Furthermore, existing multi-layer KD methods are usually unable to effectively handle networks with significantly different architectures (heterogeneous KD). In this paper we propose a novel KD method that works by modeling the information flow through the various layers of the teacher model and then train a student model to mimic this information flow. The proposed method is capable of overcoming the aforementioned limitations by using an appropriate supervision scheme during the different phases of the training process, as well as by designing and training an appropriate auxiliary teacher model that acts as a proxy model capable of "explaining" the way the teacher works to the student. The effectiveness of the proposed method is demonstrated using four image datasets and several different evaluation setups., Comment: Accepted at CVPR 2020

In this paper, we propose a novel color constancy approach, called Bag of Color Features (BoCF), building upon Bag-of-Features pooling. The proposed method substantially reduces the number of parameters needed for illumination estimation. At the same time, the proposed method is consistent with the color constancy assumption stating that global spatial information is not relevant for illumination estimation and local information ( edges, etc.) is sufficient. Furthermore, BoCF is consistent with color constancy statistical approaches and can be interpreted as a learning-based generalization of many statistical approaches. To further improve the illumination estimation accuracy, we propose a novel attention mechanism for the BoCF model with two variants based on self-attention. BoCF approach and its variants achieve competitive, compared to the state of the art, results while requiring much fewer parameters on three benchmark datasets: ColorChecker RECommended, INTEL-TUT version 2, and NUS8., Comment: 12 pages, 5 figures, 6 tables

Deep Learning (DL) models can be used to tackle time series analysis tasks with great success. However, the performance of DL models can degenerate rapidly if the data are not appropriately normalized. This issue is even more apparent when DL is used for financial time series forecasting tasks, where the non-stationary and multimodal nature of the data pose significant challenges and severely affect the performance of DL models. In this work, a simple, yet effective, neural layer, that is capable of adaptively normalizing the input time series, while taking into account the distribution of the data, is proposed. The proposed layer is trained in an end-to-end fashion using back-propagation and leads to significant performance improvements compared to other evaluated normalization schemes. The proposed method differs from traditional normalization methods since it learns how to perform normalization for a given task instead of using a fixed normalization scheme. At the same time, it can be directly applied to any new time series without requiring re-training. The effectiveness of the proposed method is demonstrated using a large-scale limit order book dataset, as well as a load forecasting dataset.

Time series forecasting is a crucial component of many important applications, ranging from forecasting the stock markets to energy load prediction. The high-dimensionality, velocity and variety of the data collected in these applications pose significant and unique challenges that must be carefully addressed for each of them. In this work, a novel Temporal Logistic Neural Bag-of-Features approach, that can be used to tackle these challenges, is proposed. The proposed method can be effectively combined with deep neural networks, leading to powerful deep learning models for time series analysis. However, combining existing BoF formulations with deep feature extractors pose significant challenges: the distribution of the input features is not stationary, tuning the hyper-parameters of the model can be especially difficult and the normalizations involved in the BoF model can cause significant instabilities during the training process. The proposed method is capable of overcoming these limitations by a employing a novel adaptive scaling mechanism and replacing the classical Gaussian-based density estimation involved in the regular BoF model with a logistic kernel. The effectiveness of the proposed approach is demonstrated using extensive experiments on a large-scale financial time series dataset that consists of more than 4 million limit orders.

Several deep supervised hashing techniques have been proposed to allow for efficiently querying large image databases. However, deep supervised image hashing techniques are developed, to a great extent, heuristically often leading to suboptimal results. Contrary to this, we propose an efficient deep supervised hashing algorithm that optimizes the learned codes using an information-theoretic measure, the Quadratic Mutual Information (QMI). The proposed method is adapted to the needs of large-scale hashing and information retrieval leading to a novel information-theoretic measure, the Quadratic Spherical Mutual Information (QSMI). Apart from demonstrating the effectiveness of the proposed method under different scenarios and outperforming existing state-of-the-art image hashing techniques, this paper provides a structured way to model the process of information retrieval and develop novel methods adapted to the needs of each application.

Universal Neural Style Transfer (NST) methods are capable of performing style transfer of arbitrary styles in a style-agnostic manner via feature transforms in (almost) real-time. Even though their unimodal parametric style modeling approach has been proven adequate to transfer a single style from relatively simple images, they are usually not capable of effectively handling more complex styles, producing significant artifacts, as well as reducing the quality of the synthesized textures in the stylized image. To overcome these limitations, in this paper we propose a novel universal NST approach that separately models each sub-style that exists in a given style image (or a collection of style images). This allows for better modeling the subtle style differences within the same style image and then using the most appropriate sub-style (or mixtures of different sub-styles) to stylize the content image. The ability of the proposed approach to a) perform a wide range of different stylizations using the sub-styles that exist in one style image, while giving the ability to the user to appropriate mix the different sub-styles, b) automatically match the most appropriate sub-style to different semantic regions of the content image, improving existing state-of-the-art universal NST approaches, and c) detecting and transferring the sub-styles from collections of images are demonstrated through extensive experiments.

Recent advances in machine learning allow us to analyze and describe the content of high-dimensional data like text, audio, images or other signals. In order to visualize that data in 2D or 3D, usually Dimensionality Reduction (DR) techniques are employed. Most of these techniques, e.g., PCA or t-SNE, produce static projections without taking into account corrections from humans or other data exploration scenarios. In this work, we propose the interactive Similarity Projection (iSP), a novel interactive DR framework based on similarity embeddings, where we form a differentiable objective based on the user interactions and perform learning using gradient descent, with an end-to-end trainable architecture. Two interaction scenarios are evaluated. First, a common methodology in multidimensional projection is to project a subset of data, arrange them in classes or clusters, and project the rest unseen dataset based on that manipulation, in a kind of semi-supervised interpolation. We report results that outperform competitive baselines in a wide range of metrics and datasets. Second, we explore the scenario of manipulating some classes, while enriching the optimization with high-dimensional neighbor information. Apart from improving classification precision and clustering on images and text documents, the new emerging structure of the projection unveils semantic manifolds. For example, on the Head Pose dataset, by just dragging the faces looking far left to the left and those looking far right to the right, all faces are re-arranged on a continuum even on the vertical axis (face up and down). This end-to-end framework can be used for fast, visual semi-supervised learning, manifold exploration, interactive domain adaptation of neural embeddings and transfer learning., Comment: Accepted at Knowledge-Based Systems

Among the most impressive recent applications of neural decoding is the visual representation decoding, where the category of an object that a subject either sees or imagines is inferred by observing his/her brain activity. Even though there is an increasing interest in the aforementioned visual representation decoding task, there is no extensive study of the effect of using different machine learning models on the decoding accuracy. In this paper we provide an extensive evaluation of several machine learning models, along with different similarity metrics, for the aforementioned task, drawing many interesting conclusions. That way, this paper a) paves the way for developing more advanced and accurate methods and b) provides an extensive and easily reproducible baseline for the aforementioned decoding task., Comment: Accepted at 1st Workshop on Brain-Driven Computer Vision - ECCV 2018

The recent surge in Deep Learning (DL) research of the past decade has successfully provided solutions to many difficult problems. The field of quantitative analysis has been slowly adapting the new methods to its problems, but due to problems such as the non-stationary nature of financial data, significant challenges must be overcome before DL is fully utilized. In this work a new method to construct stationary features, that allows DL models to be applied effectively, is proposed. These features are thoroughly tested on the task of predicting mid price movements of the Limit Order Book. Several DL models are evaluated, such as recurrent Long Short Term Memory (LSTM) networks and Convolutional Neural Networks (CNN). Finally a novel model that combines the ability of CNNs to extract useful features and the ability of LSTMs' to analyze time series, is proposed and evaluated. The combined model is able to outperform the individual LSTM and CNN models in the prediction horizons that are tested.