With the rapid development of in-depth learning, neural network and deep learning algorithms have been widely used in various fields, e.g., image, video and voice processing. However, the neural network model is getting larger and larger, which is expressed in the calculation of model parameters. Although a wealth of existing efforts on GPU platforms currently used by researchers for improving computing performance, dedicated hardware solutions are essential and emerging to provide advantages over pure software solutions. In this paper, we systematically investigate the neural network accelerator based on FPGA. Specifically, we respectively review the accelerators designed for specific problems, specific algorithms, algorithm features, and general templates. We also compared the design and implementation of the accelerator based on FPGA under different devices and network models and compared it with the versions of CPU and GPU. Finally, we present to discuss the advantages and disadvantages of accelerators on FPGA platforms and to further explore the opportunities for future research.
The popularity of Deep Learning for real-world applications is ever-growing. With the introduction of high performance hardware, applications are no longer limited to image recognition. With the introduction of more complex problems comes more and more complex solutions, and the increasing need for explainable AI. Deep Neural Networks for Video tasks are amongst the most complex models, with at least twice the parameters of their Image counterparts. However, explanations for these models are often ill-adapted to the video domain. The current work in explainability for video models is still overshadowed by Image techniques, while Video Deep Learning itself is quickly gaining on methods for still images. This paper seeks to highlight the need for explainability methods designed with video deep learning models, and by association spatio-temporal input in mind, by first illustrating the cutting edge for video deep learning, and then noting the scarcity of research into explanations for these methods.
Text document classification is an important task for diverse natural language processing based applications. Traditional machine learning approaches mainly focused on reducing dimensionality of textual data to perform classification. This although improved the overall classification accuracy, the classifiers still faced sparsity problem due to lack of better data representation techniques. Deep learning based text document classification, on the other hand, benefitted greatly from the invention of word embeddings that have solved the sparsity problem and researchers focus mainly remained on the development of deep architectures. Deeper architectures, however, learn some redundant features that limit the performance of deep learning based solutions. In this paper, we propose a two stage text document classification methodology which combines traditional feature engineering with automatic feature engineering (using deep learning). The proposed methodology comprises a filter based feature selection (FSE) algorithm followed by a deep convolutional neural network. This methodology is evaluated on the two most commonly used public datasets, i.e., 20 Newsgroups data and BBC news data. Evaluation results reveal that the proposed methodology outperforms the state-of-the-art of both the (traditional) machine learning and deep learning based text document classification methodologies with a significant margin of 7.7% on 20 Newsgroups and 6.6% on BBC news datasets.
Skin conditions affect an estimated 1.9 billion people worldwide. A shortage of dermatologists causes long wait times and leads patients to seek dermatologic care from general practitioners. However, the diagnostic accuracy of general practitioners has been reported to be only 0.24-0.70 (compared to 0.77-0.96 for dermatologists), resulting in referral errors, delays in care, and errors in diagnosis and treatment. In this paper, we developed a deep learning system (DLS) to provide a differential diagnosis of skin conditions for clinical cases (skin photographs and associated medical histories). The DLS distinguishes between 26 skin conditions that represent roughly 80% of the volume of skin conditions seen in primary care. The DLS was developed and validated using de-identified cases from a teledermatology practice serving 17 clinical sites via a temporal split: the first 14,021 cases for development and the last 3,756 cases for validation. On the validation set, where a panel of three board-certified dermatologists defined the reference standard for every case, the DLS achieved 0.71 and 0.93 top-1 and top-3 accuracies respectively. For a random subset of the validation set (n=963 cases), 18 clinicians reviewed the cases for comparison. On this subset, the DLS achieved a 0.67 top-1 accuracy, non-inferior to board-certified dermatologists (0.63, p<0.001), and higher than primary care physicians (PCPs, 0.45) and nurse practitioners (NPs, 0.41). The top-3 accuracy showed a similar trend: 0.90 DLS, 0.75 dermatologists, 0.60 PCPs, and 0.55 NPs. These results highlight the potential of the DLS to augment general practitioners to accurately diagnose skin conditions by suggesting differential diagnoses that may not have been considered. Future work will be needed to prospectively assess the clinical impact of using this tool in actual clinical workflows.
Deep neural networks (DNNs) provide high image classification accuracy, but experience significant performance degradation when perturbation from various sources are present in the input. The lack of resilience to input perturbations makes DNN less reliable for systems interacting with physical world such as autonomous vehicles, robotics, to name a few, where imperfect input is the normal condition. We present a hybrid deep network architecture with spike-assisted contextual information extraction (ScieNet). ScieNet integrates unsupervised learning using spiking neural network (SNN) for unsupervised contextual informationextraction with a back-end DNN trained for classification. The integrated network demonstrates high resilience to input perturbations without relying on prior training on perturbed inputs. We demonstrate ScieNet with different back-end DNNs for image classification using CIFAR dataset considering stochastic (noise) and structured (rain) input perturbations. Experimental results demonstrate significant improvement in accuracy on noisy and rainy images without prior training, while maintaining state-of-the-art accuracy on clean images.
With the rise of social media like Twitter and of software distribution platforms like app stores, users got various ways to express their opinion about software products. Popular software vendors get user feedback thousandfold per day. Research has shown that such feedback contains valuable information for software development teams such as problem reports or feature and support inquires. Since the manual analysis of user feedback is cumbersome and hard to manage many researchers and tool vendors suggested to use automated analyses based on traditional supervised machine learning approaches. In this work, we compare the results of traditional machine learning and deep learning in classifying user feedback in English and Italian into problem reports, inquiries, and irrelevant. Our results show that using traditional machine learning, we can still achieve comparable results to deep learning, although we collected thousands of labels.
We develop a Distributed Event-Triggered Stochastic GRAdient Descent (DETSGRAD) algorithm for solving non-convex optimization problems typically encountered in distributed deep learning. We propose a novel communication triggering mechanism that would allow the networked agents to update their model parameters aperiodically and provide sufficient conditions on the algorithm step-sizes that guarantee the asymptotic mean-square convergence. The algorithm is applied to a distributed supervised-learning problem, in which a set of networked agents collaboratively train their individual neural networks to recognize handwritten digits in images, while aperiodically sharing the model parameters with their one-hop neighbors. Results indicate that all agents report similar performance that is also comparable to the performance of a centrally trained neural network, while the event-triggered communication provides significant reduction in inter-agent communication. Results also show that the proposed algorithm allows the individual agents to recognize the digits even though the training data corresponding to all the digits are not locally available to each agent.
Reinforcement learning has exceeded human-level performance in game playing AI with deep learning methods according to the experiments from DeepMind on Go and Atari games. Deep learning solves high dimension input problems which stop the development of reinforcement for many years. This study uses both two techniques to create several agents with different algorithms that successfully learn to play T-rex Runner. Deep Q network algorithm and three types of improvements are implemented to train the agent. The results from some of them are far from satisfactory but others are better than human experts. Batch normalization is a method to solve internal covariate shift problems in deep neural network. The positive influence of this on reinforcement learning has also been proved in this study.
Regularization methods are often employed in deep learning neural networks (DNNs) to prevent overfitting. For penalty based DNN regularization methods, convex penalties are typically considered because of their optimization guarantees. Recent theoretical work have shown that nonconvex penalties that satisfy certain regularity conditions are also guaranteed to perform well with standard optimization algorithms. In this paper, we examine new and currently existing nonconvex penalties for DNN regularization. We provide theoretical justifications for the new penalties and also assess the performance of all penalties with DNN analyses of seven datasets.
Deep Networks have been shown to provide state-of-the-art performance in many machine learning challenges. Unfortunately, they are susceptible to various types of noise, including adversarial attacks and corrupted inputs. In this work we introduce a formal definition of robustness which can be viewed as a localized Lipschitz constant of the network function, quantified in the domain of the data to be classified. We compare this notion of robustness to existing ones, and study its connections with methods in the literature. We evaluate this metric by performing experiments on various competitive vision datasets.
This paper brings deep learning at the forefront of research into Time Series Classification (TSC). TSC is the area of machine learning tasked with the categorization (or labelling) of time series. The last few decades of work in this area have led to significant progress in the accuracy of classifiers, with the state of the art now represented by the HIVE-COTE algorithm. While extremely accurate, HIVE-COTE cannot be applied to many real-world datasets because of its high training time complexity in O(N2 * T4) for a dataset with N time series of length T. For example, it takes HIVE-COTE more than 8 days to learn from a small dataset with N = 1500 time series of short length T = 46. Meanwhile deep learning has received enormous attention because of its high accuracy and scalability. Recent approaches to deep learning for TSC have been scalable, but less accurate than HIVE-COTE. We introduce InceptionTime - an ensemble of deep Convolutional Neural Network (CNN) models, inspired by the Inception-v4 architecture. Our experiments show that InceptionTime is on par with HIVE-COTE in terms of accuracy while being much more scalable: not only can it learn from 1,500 time series in one hour but it can also learn from 8M time series in 13 hours, a quantity of data that is fully out of reach of HIVE-COTE.
We study an emerging problem named "grasping the invisible" in robotic manipulation, in which a robot is tasked to grasp an initially invisible target object via a sequence of pushing and grasping actions. In this problem, pushes are needed to search for the target and rearrange cluttered objects around it to enable effective grasps. We propose to solve the problem by formulating a deep learning approach in a critic-policy format. The target-oriented motion critic, which maps both visual observations and target information to the expected future rewards of pushing and grasping motion primitives, is learned via deep Q-learning. We divide the problem into two subtasks, and two policies are proposed to tackle each of them, by combining the critic predictions and relevant domain knowledge. A Bayesian-based policy accounting for past action experience performs pushing to search for the target; once the target is found, a classifier-based policy coordinates target-oriented pushing and grasping to grasp the target in clutter. The motion critic and the classifier are trained in a self-supervised manner through robot-environment interactions. Our system achieves a 93% and 87% task success rate on each of the two subtasks in simulation and an 85% task success rate in real robot experiments on the whole problem, which outperforms several baselines by large margins. Supplementary material is available at https://sites.google.com/umn.edu/grasping-invisible.
Deep neural networks have introduced novel and useful tools to the machine learning community. Other types of classifiers can potentially make use of these tools as well to improve their performance and generality. This paper reviews the current state of the art for deep learning classifier technologies that are being used outside of deep neural networks. Non-network classifiers can employ many components found in deep neural network architectures. In this paper, we review the feature learning, optimization, and regularization methods that form a core of deep network technologies. We then survey non-neural network learning algorithms that make innovative use of these methods to improve classification. Because many opportunities and challenges still exist, we discuss directions that can be pursued to expand the area of deep learning for a variety of classification algorithms.
The tremendous success of deep learning for imaging applications has resulted in numerous beneficial advances. Unfortunately, this success has also been a catalyst for malicious uses such as photo-realistic face swapping of parties without consent. Transferring one person's face from a source image to a target image of another person, while keeping the image photo-realistic overall has become increasingly easy and automatic, even for individuals without much knowledge of image processing. In this study, we use deep transfer learning for face swapping detection, showing true positive rates >96% with very few false alarms. Distinguished from existing methods that only provide detection accuracy, we also provide uncertainty for each prediction, which is critical for trust in the deployment of such detection systems. Moreover, we provide a comparison to human subjects. To capture human recognition performance, we build a website to collect pairwise comparisons of images from human subjects. Based on these comparisons, images are ranked from most real to most fake. We compare this ranking to the outputs from our automatic model, showing good, but imperfect, correspondence with linear correlations >0.75. Overall, the results show the effectiveness of our method. As part of this study, we create a novel, publicly available dataset that is, to the best of our knowledge, the largest public swapped face dataset created using still images. Our goal of this study is to inspire more research in the field of image forensics through the creation of a public dataset and initial analysis.
In high population cities, the gatherings of large crowds in public places and public areas accelerate or jeopardize people safety and transportation, which is a key challenge to the researchers. Although much research has been carried out on crowd analytics, many of existing methods are problem-specific, i.e., methods learned from a specific scene cannot be properly adopted to other videos. Therefore, this presents weakness and the discovery of these researches, since additional training samples have to be found from diverse videos. This paper will investigate diverse scene crowd analytics with traditional and deep learning models. We will also consider pros and cons of these approaches. However, once general deep methods are investigated from large datasets, they can be consider to investigate different crowd videos and images. Therefore, it would be able to cope with the problem including to not limited to crowd density estimation, crowd people counting, and crowd event recognition. Deep learning models and approaches are required to have large datasets for training and testing. Many datasets are collected taking into account many different and various problems related to building crowd datasets, including manual annotations and increasing diversity of videos and images. In this paper, we will also propose many models of deep neural networks and training approaches to learn the feature modeling for crowd analytics.
The use of semantic segmentation for masking and cropping input images has proven to be a significant aid in medical imaging classification tasks by decreasing the noise and variance of the training dataset. However, implementing this approach with classical methods is challenging: the cost of obtaining a dense segmentation is high, and the precise input area that is most crucial to the classification task is difficult to determine a-priori. We propose a novel joint-training deep reinforcement learning framework for image augmentation. A segmentation network, weakly supervised with policy gradient optimization, acts as an agent, and outputs masks as actions given samples as states, with the goal of maximizing reward signals from the classification network. In this way, the segmentation network learns to mask unimportant imaging features. Our method, Adversarial Policy Gradient Augmentation (APGA), shows promising results on Stanford's MURA dataset and on a hip fracture classification task with an increase in global accuracy of up to 7.33% and improved performance over baseline methods in 9/10 tasks evaluated. We discuss the broad applicability of our joint training strategy to a variety of medical imaging tasks.
It is difficult to quantify structure-property relationships and to identify structural features of complex materials. The characterization of amorphous materials is especially challenging because their lack of long-range order makes it difficult to define structural metrics. In this work, we apply deep learning algorithms to accurately classify amorphous materials and characterize their structural features. Specifically, we show that convolutional neural networks and message passing neural networks can classify two-dimensional liquids and liquid-cooled glasses from molecular dynamics simulations with greater than 0.98 AUC, with no a priori assumptions about local particle relationships, even when the liquids and glasses are prepared at the same inherent structure energy. Furthermore, we demonstrate that message passing neural networks surpass convolutional neural networks in this context in both accuracy and interpretability. We extract a clear interpretation of how message passing neural networks evaluate liquid and glass structures by using a self-attention mechanism. Using this interpretation, we derive three novel structural metrics that accurately characterize glass formation. The methods presented here provide us with a procedure to identify important structural features in materials that could be missed by standard techniques and give us a unique insight into how these neural networks process data.
We present a study of deep learning applied to the domain of network traffic data forecasting. This is a very important ingredient for network traffic engineering, e.g., intelligent routing, which can optimize network performance, especially in large networks. In a nutshell, we wish to predict, in advance, the bit rate for a transmission, based on low-dimensional connection metadata ("flows") that is available whenever a communication is initiated. Our study has several genuinely new points: First, it is performed on a large dataset (~50 million flows), which requires a new training scheme that operates on successive blocks of data since the whole dataset is too large for in-memory processing. Additionally, we are the first to propose and perform a more fine-grained prediction that distinguishes between low, medium and high bit rates instead of just "mice" and "elephant" flows. Lastly, we apply state-of-the-art visualization and clustering techniques to flow data and show that visualizations are insightful despite the heterogeneous and non-metric nature of the data. We developed a processing pipeline to handle the highly non-trivial acquisition process and allow for proper data preprocessing to be able to apply DNNs to network traffic data. We conduct DNN hyper-parameter optimization as well as feature selection experiments, which clearly show that fine-grained network traffic forecasting is feasible, and that domain-dependent data enrichment and augmentation strategies can improve results. An outlook about the fundamental challenges presented by network traffic analysis (high data throughput, unbalanced and dynamic classes, changing statistics, outlier detection) concludes the article.
Reliability is a critical consideration to DL-based systems. But the statistical nature of DL makes it quite vulnerable to invalid inputs, i.e., those cases that are not considered in the training phase of a DL model. This paper proposes to perform data sanity check to identify invalid inputs, so as to enhance the reliability of DL-based systems. We design and implement a tool to detect behavior deviation of a DL model when processing an input case. This tool extracts the data flow footprints and conducts an assertion-based validation mechanism. The assertions are built automatically, which are specifically-tailored for DL model data flow analysis. Our experiments conducted with real-world scenarios demonstrate that such an assertion-based data sanity check mechanism is effective in identifying invalid input cases.
Supervised deep learning requires large amounts of training data. In the context of the FIRE2019 Arabic irony detection shared task (IDAT@FIRE2019), we show how we mitigate this need by fine-tuning the pre-trained bidirectional encoders from transformers (BERT) on gold data in a multi-task setting. We further improve our models by by further pre-training BERT on `in-domain' data, thus alleviating an issue of dialect mismatch in the Google-released BERT model. Our best model acquires 82.4 macro F1 score, and has the unique advantage of being feature-engineering free (i.e., based exclusively on deep learning).