With the rapid advancements of deep learning in the past decade, it can be foreseen that deep learning will be continuously deployed in more and more safety-critical applications such as autonomous driving and robotics. In this context, reliability turns out to be critical to the deployment of deep learning in these applications and gradually becomes a first-class citizen among the major design metrics like performance and energy efficiency. Nevertheless, the back-box deep learning models combined with the diverse underlying hardware faults make resilient deep learning extremely challenging. In this special session, we conduct a comprehensive survey of fault-tolerant deep learning design approaches with a hierarchical perspective and investigate these approaches from model layer, architecture layer, circuit layer, and cross layer respectively.
Agricultural datasets for crop row detection are often bound by their limited number of images. This restricts the researchers from developing deep learning based models for precision agricultural tasks involving crop row detection. We suggest the utilization of small real-world datasets along with additional data generated by simulations to yield similar crop row detection performance as that of a model trained with a large real world dataset. Our method could reach the performance of a deep learning based crop row detection model trained with real-world data by using 60% less labelled real-world data. Our model performed well against field variations such as shadows, sunlight and grow stages. We introduce an automated pipeline to generate labelled images for crop row detection in simulation domain. An extensive comparison is done to analyze the contribution of simulated data towards reaching robust crop row detection in various real-world field scenarios.
The deep learning revolution incited by the 2012 Alexnet paper has been transformative for the field of computer vision. Many problems which were severely limited using classical solutions are now seeing unprecedented success. The rapid proliferation of deep learning methods has led to a sharp increase in their use in consumer and embedded applications. One consequence of consumer and embedded applications is lossy multimedia compression which is required to engineer the efficient storage and transmission of data in these real-world scenarios. As such, there has been increased interest in a deep learning solution for multimedia compression which would allow for higher compression ratios and increased visual quality. The deep learning approach to multimedia compression, so called Learned Multimedia Compression, involves computing a compressed representation of an image or video using a deep network for the encoder and the decoder. While these techniques have enjoyed impressive academic success, their industry adoption has been essentially non-existent. Classical compression techniques like JPEG and MPEG are too entrenched in modern computing to be easily replaced. This dissertation takes an orthogonal approach and leverages deep learning to improve the compression fidelity of these classical algorithms. This allows the incredible advances in deep learning to be used for multimedia compression without threatening the ubiquity of the classical methods. The key insight of this work is that methods which are motivated by first principles, i.e., the underlying engineering decisions that were made when the compression algorithms were developed, are more effective than general methods. By encoding prior knowledge into the design of the algorithm, the flexibility, performance, and/or accuracy are improved at the cost of generality...
The vast majority of systems of practical interest are characterised by nonlinear dynamics. This renders the control and optimization of such systems a complex task due to their nonlinear behaviour. Additionally, standard methods such as linearizing around a fixed point may not be an effective strategy for many systems, thus requiring an alternative approach. For this reason, we propose a new deep learning framework to discover a transformation of a nonlinear dynamical system to an equivalent higher dimensional linear representation. We demonstrate that the resulting learned linear representation accurately captures the dynamics of the original system for a wider range of conditions than standard linearization. As a result of this, we show that the learned linear model can subsequently be used for the successful control of the original system. We demonstrate this by applying the proposed framework to two examples; one from the literature and a more complex example in the form of a Continuous Stirred Tank Reactor (CSTR).
Due to the Internet of Things (IoT) proliferation, Radio Frequency (RF) channels are increasingly congested with new kinds of devices, which carry unique and diverse communication needs. This poses complex challenges in modern digital communications, and calls for the development of technological innovations that (i) optimize capacity (bitrate) in limited bandwidth environments, (ii) integrate cooperatively with already-deployed RF protocols, and (iii) are adaptive to the ever-changing demands in modern digital communications. In this paper we present methods for applying deep neural networks for spectral filling. Given an RF channel transmitting digital messages with a pre-established modulation scheme, we automatically learn novel modulation schemes for sending extra information, in the form of additional messages, "around" the fixed-modulation signals (i.e., without interfering with them). In so doing, we effectively increase channel capacity without increasing bandwidth. We further demonstrate the ability to generate signals that closely resemble the original modulations, such that the presence of extra messages is undetectable to third-party listeners. We present three computational experiments demonstrating the efficacy of our methods, and conclude by discussing the implications of our results for modern RF applications.
Stochastic gradient descent (SGD) is widely used in deep learning due to its computational efficiency, but a complete understanding of why SGD performs so well remains a major challenge. It has been observed empirically that most eigenvalues of the Hessian of the loss functions on the loss landscape of over-parametrized deep neural networks are close to zero, while only a small number of eigenvalues are large. Zero eigenvalues indicate zero diffusion along the corresponding directions. This indicates that the process of minima selection mainly happens in the relatively low-dimensional subspace corresponding to the top eigenvalues of the Hessian. Although the parameter space is very high-dimensional, these findings seems to indicate that the SGD dynamics may mainly live on a low-dimensional manifold. In this paper, we pursue a truly data driven approach to the problem of getting a potentially deeper understanding of the high-dimensional parameter surface, and in particular, of the landscape traced out by SGD by analyzing the data generated through SGD, or any other optimizer for that matter, in order to possibly discover (local) low-dimensional representations of the optimization landscape. As our vehicle for the exploration, we use diffusion maps introduced by R. Coifman and coauthors.
The early detection of drowsiness has become vital to ensure the correct and safe development of several industries' tasks. Due to the transient mental state of a human subject between alertness and drowsiness, automated drowsiness detection is a complex problem to tackle. The electroencephalography signals allow us to record variations in an individual's brain's electrical potential, where each of them gives specific information about a subject's mental state. However, due to this type of signal's nature, its acquisition, in general, is complex, so it is hard to have a large volume of data to apply techniques of Deep Learning for processing and classification optimally. Nevertheless, Capsule Neural Networks are a brand-new Deep Learning algorithm proposed for work with reduced amounts of data. It is a robust algorithm to handle the data's hierarchical relationships, which is an essential characteristic for work with biomedical signals. Therefore, this paper presents a Deep Learning-based method for drowsiness detection with CapsNet by using a concatenation of spectrogram images of the electroencephalography signals channels. The proposed CapsNet model is compared with a Convolutional Neural Network, which is outperformed by the proposed model, which obtains an average accuracy of 86,44% and 87,57% of sensitivity against an average accuracy of 75,86% and 79,47% sensitivity for the CNN, showing that CapsNet is more suitable for this kind of datasets and tasks.
Airports have an important role in both military and civilian domains. The synthetic aperture radar (SAR) based airport detection has received increasing attention in recent years. However, due to the high cost of SAR imaging and annotation process, there is no publicly available SAR dataset for airport detection. As a result, deep learning methods have not been fully used in airport detection tasks. To provide a benchmark for airport detection research in SAR images, this paper introduces a large-scale SAR Airport Dataset (SAD). In order to adequately reflect the demands of real world applications, it contains 624 SAR images from Sentinel 1B and covers 104 airfield instances with different scales, orientations and shapes. The experiments of multiple deep learning approach on this dataset proves its effectiveness. It developing state-of-the-art airport area detection algorithms or other relevant tasks.
Spiking Neural Networks (SNNs) have attracted the attention of the deep learning community for use in low-latency, low-power neuromorphic hardware, as well as models for understanding neuroscience. In this paper, we introduce Spiking Phasor Neural Networks (SPNNs). SPNNs are based on complex-valued Deep Neural Networks (DNNs), representing phases by spike times. Our model computes robustly employing a spike timing code and gradients can be formed using the complex domain. We train SPNNs on CIFAR-10, and demonstrate that the performance exceeds that of other timing coded SNNs, approaching results with comparable real-valued DNNs.
Machine learning (ML) and a nonlinear model predictive controller (NMPC) are used in this paper to minimize the emissions and fuel consumption of a compression ignition engine. In this work machine learning is applied in two methods. In the first application, ML is used to identify a model for implementation in model predictive control optimization problems. In the second application, ML is used as a replacement of the NMPC where the ML controller learns the optimal control action by imitating or mimicking the behavior of the model predictive controller. In this study, a deep recurrent neural network including long-short term memory (LSTM) layers are used to model the emissions and performance of an industrial 4.5 liter 4-cylinder Cummins diesel engine. This model is then used for model predictive controller implementation. Then, a deep learning scheme is deployed to clone the behavior of the developed controller. In the LSTM integration, a novel scheme is used by augmenting hidden and cell states of the network in an NMPC optimization problem. The developed LSTM-NMPC and the imitative NMPC are compared with the Cummins calibrated Engine Control Unit (ECU) model in an experimentally validated engine simulation platform. Results show a significant reduction in Nitrogen Oxides (\nox) emissions and a slight decrease in the injected fuel quantity while maintaining the same load. In addition, the imitative NMPC has a similar performance as the NMPC but with a two orders of magnitude reduction of the computation time.
Weeds are a major threat to crops and are responsible for reducing crop yield worldwide. To mitigate their negative effect, it is advantageous to accurately identify them early in the season to prevent their spread throughout the field. Traditionally, farmers rely on manually scouting fields and applying herbicides for different weeds. However, it is easy to confuse between crops with weeds during the early growth stages. Recently, deep learning-based weed identification has become popular as deep learning relies on convolutional neural networks that are capable of learning important distinguishable features between weeds and crops. However, training robust deep learning models requires access to large imagery datasets. Therefore, an early-season weeds dataset was acquired under field conditions. The dataset consists of 159 Cocklebur images, 139 Foxtail images, 170 Redroot Pigweed images and 150 Giant Ragweed images corresponding to four common weed species found in corn and soybean production systems.. Bounding box annotations were created for each image to prepare the dataset for training both image classification and object detection deep learning networks capable of accurately locating and identifying weeds within corn and soybean fields. (https://osf.io/w9v3j/)
Alzheimer's disease (AD) is one of the most common public health issues the world is facing today. This disease has a high prevalence primarily in the elderly accompanying memory loss and cognitive decline. AD detection is a challenging task which many authors have developed numerous computerized automatic diagnosis systems utilizing neuroimaging and other clinical data. MRI scans provide high-intensity visible features, making these scans the most widely used brain imaging technique. In recent years deep learning has achieved leading success in medical image analysis. But a relatively little investigation has been done to apply deep learning techniques for the brain MRI classification. This paper explores the construction of several deep learning architectures evaluated on brain MRI images and segmented images. The idea behind segmented images investigates the influence of image segmentation step on deep learning classification. The image processing presented a pipeline consisting of pre-processing to enhance the MRI scans and post-processing consisting of a segmentation method for segmenting the brain tissues. The results show that the processed images achieved a better accuracy in the binary classification of AD vs. CN (Cognitively Normal) across four different architectures. ResNet architecture resulted in the highest prediction accuracy amongst the other architectures (90.83% for the original brain images and 93.50% for the processed images).
Lung cancer is the leading cause of cancer-related mortality worldwide. Lung cancer screening (LCS) using annual low-dose computed tomography (CT) scanning has been proven to significantly reduce lung cancer mortality by detecting cancerous lung nodules at an earlier stage. Improving risk stratification of malignancy risk in lung nodules can be enhanced using machine/deep learning algorithms. However most existing algorithms: a) have primarily assessed single time-point CT data alone thereby failing to utilize the inherent advantages contained within longitudinal imaging datasets; b) have not integrated into computer models pertinent clinical data that might inform risk prediction; c) have not assessed algorithm performance on the spectrum of nodules that are most challenging for radiologists to interpret and where assistance from analytic tools would be most beneficial. Here we show the performance of our time-series deep learning model (DeepCAD-NLM-L) which integrates multi-model information across three longitudinal data domains: nodule-specific, lung-specific, and clinical demographic data. We compared our time-series deep learning model to a) radiologist performance on CTs from the National Lung Screening Trial enriched with the most challenging nodules for diagnosis; b) a nodule management algorithm from a North London LCS study (SUMMIT). Our model demonstrated comparable and complementary performance to radiologists when interpreting challenging lung nodules and showed improved performance (AUC=88\%) against models utilizing single time-point data only. The results emphasise the importance of time-series, multi-modal analysis when interpreting malignancy risk in LCS.
Orbital angular momentum of light is regarded as a valuable resource in quantum technology, especially in quantum communication and quantum sensing and ranging. However, the OAM state of light is susceptible to undesirable experimental conditions such as propagation distance and phase distortions, which hinders the potential for the realistic implementation of relevant technologies. In this article, we exploit an enhanced deep learning neural network to identify different OAM modes of light at multiple propagation distances with phase distortions. Specifically, our trained deep learning neural network can efficiently identify the vortex beam's topological charge and propagation distance with 97% accuracy. Our technique has important implications for OAM based communication and sensing protocols.
Theory of Mind is an essential ability of humans to infer the mental states of others. Here we provide a coherent summary of the potential, current progress, and problems of deep learning approaches to Theory of Mind. We highlight that many current findings can be explained through shortcuts. These shortcuts arise because the tasks used to investigate Theory of Mind in deep learning systems have been too narrow. Thus, we encourage researchers to investigate Theory of Mind in complex open-ended environments. Furthermore, to inspire future deep learning systems we provide a concise overview of prior work done in humans. We further argue that when studying Theory of Mind with deep learning, the research's main focus and contribution ought to be opening up the network's representations. We recommend researchers use tools from the field of interpretability of AI to study the relationship between different network components and aspects of Theory of Mind.
In this paper, we explore the possibility of detecting polar lows in C-band SAR images by means of deep learning. Specifically, we introduce a novel dataset consisting of Sentinel-1 images divided into two classes, representing the presence and absence of a maritime mesocyclone, respectively. The dataset is constructed using the ERA5 dataset as baseline and it consists of 2004 annotated images. To our knowledge, this is the first dataset of its kind to be publicly released. The dataset is used to train a deep learning model to classify the labeled images. Evaluated on an independent test set, the model yields an F-1 score of 0.95, indicating that polar lows can be consistently detected from SAR images. Interpretability techniques applied to the deep learning model reveal that atmospheric fronts and cyclonic eyes are key features in the classification. Moreover, experimental results show that the model is accurate even if: (i) such features are significantly cropped due to the limited swath width of the SAR, (ii) the features are partly covered by sea ice and (iii) land is covering significant parts of the images. By evaluating the model performance on multiple input image resolutions (pixel sizes of 500m, 1km and 2km), it is found that higher resolution yield the best performance. This emphasises the potential of using high resolution sensors like SAR for detecting polar lows, as compared to conventionally used sensors such as scatterometers.
Autonomous driving technology has drawn a lot of attention due to its fast development and extremely high commercial values. The recent technological leap of autonomous driving can be primarily attributed to the progress in the environment perception. Good environment perception provides accurate high-level environment information which is essential for autonomous vehicles to make safe and precise driving decisions and strategies. Moreover, such progress in accurate environment perception would not be possible without deep learning models and advanced onboard sensors, such as optical sensors (LiDARs and cameras), radars, GPS. However, the advanced sensors and deep learning models are prone to recently invented attack methods. For example, LiDARs and cameras can be compromised by optical attacks, and deep learning models can be attacked by adversarial examples. The attacks on advanced sensors and deep learning models can largely impact the accuracy of the environment perception, posing great threats to the safety and security of autonomous vehicles. In this thesis, we study the detection methods against the attacks on onboard sensors and the linkage between attacked deep learning models and driving safety for autonomous vehicles. To detect the attacks, redundant data sources can be exploited, since information distortions caused by attacks in victim sensor data result in inconsistency with the information from other redundant sources. To study the linkage between attacked deep learning models and driving safety...
Deep-learning-based compressor has received interests recently due to much improved compression ratio. However, modern approaches suffer from long execution time. To ease this problem, this paper targets on cutting down the execution time of deep-learning-based compressors. Building history-dependencies sequentially (e.g., recurrent neural networks) is responsible for long inference latency. Instead, we introduce transformer into deep learning compressors to build history-dependencies in parallel. However, existing transformer is too heavy in computation and incompatible to compression tasks. This paper proposes a fast general-purpose lossless compressor, TRACE, by designing a compression-friendly structure based on a single-layer transformer. We first design a new metric to advise the selection part of compression model structures. Byte-grouping and Shared-ffn schemes are further proposed to fully utilize the capacity of the single-layer transformer. These features allow TRACE to achieve competitive compression ratio and a much faster speed. In addition, we further accelerate the compression procedure by designing a controller to reduce the parameter updating overhead. Experiments show that TRACE achieves an overall $\sim$3x speedup while keeps a comparable compression ratio to the state-of-the-art compressors. The source code for TRACE and links to the datasets are available at https://github.com/mynotwo/A-Fast-Transformer-based-General-Purpose-LosslessCompressor.
In this study, the problem of computing a sparse representation of multi-dimensional visual data is considered. In general, such data e.g., hyperspectral images, color images or video data consists of signals that exhibit strong local dependencies. A new computationally efficient sparse coding optimization problem is derived by employing regularization terms that are adapted to the properties of the signals of interest. Exploiting the merits of the learnable regularization techniques, a neural network is employed to act as structure prior and reveal the underlying signal dependencies. To solve the optimization problem Deep unrolling and Deep equilibrium based algorithms are developed, forming highly interpretable and concise deep-learning-based architectures, that process the input dataset in a block-by-block fashion. Extensive simulation results, in the context of hyperspectral image denoising, are provided, which demonstrate that the proposed algorithms outperform significantly other sparse coding approaches and exhibit superior performance against recent state-of-the-art deep-learning-based denoising models. In a wider perspective, our work provides a unique bridge between a classic approach, that is the sparse representation theory, and modern representation tools that are based on deep learning modeling.
Since 2016, deep learning (DL) has advanced tomographic imaging with remarkable successes, especially in low-dose computed tomography (LDCT) imaging. Despite being driven by big data, the LDCT denoising and pure end-to-end reconstruction networks often suffer from the black box nature and major issues such as instabilities, which is a major barrier to apply deep learning methods in low-dose CT applications. An emerging trend is to integrate imaging physics and model into deep networks, enabling a hybridization of physics/model-based and data-driven elements. %This type of hybrid methods has become increasingly influential. In this paper, we systematically review the physics/model-based data-driven methods for LDCT, summarize the loss functions and training strategies, evaluate the performance of different methods, and discuss relevant issues and future directions.