Stuttering, also called stammering, is a communication disorder that breaks the continuity of the speech. This program of work is an attempt to develop automatic recognition procedures to assess stuttered dysfluencies and use these assessments to filter out speech therapies for an individual. Stuttering may be in the form of repetitions, prolongations or abnormal stoppages of sounds and syllables. Our system aims to help stutterers by diagnosing the severity and type of stutter and also by suggesting appropriate therapies for practice by learning the correlation between stutter descriptors and the effectiveness of speech therapies on them. This paper focuses on the implementation of a stutter diagnosis agent using Gated Recurrent CNN on MFCC audio features and therapy recommendation agent using SVM. It also presents the results obtained and various key findings of the system developed.
Diabetic Foot Ulcers (DFU) that affect the lower extremities are a major complication of diabetes. Each year, more than 1 million diabetic patients undergo amputation due to failure to recognize DFU and get the proper treatment from clinicians. There is an urgent need to use a CAD system for the detection of DFU. In this paper, we propose using deep learning methods (EfficientDet Architectures) for the detection of DFU in the DFUC2020 challenge dataset, which consists of 4,500 DFU images. We further refined the EfficientDet architecture to avoid false negative and false positive predictions. The code for this method is available at https://github.com/Manugoyal12345/Yet-Another-EfficientDet-Pytorch.
The geometric structure of an optimization landscape is argued to be fundamentally important to support the success of deep neural network learning. A direct computation of the landscape beyond two layers is hard. Therefore, to capture the global view of the landscape, an interpretable model of the network-parameter (or weight) space must be established. However, the model is lacking so far. Furthermore, it remains unknown what the landscape looks like for deep networks of binary synapses, which plays a key role in robust and energy efficient neuromorphic computation. Here, we propose a statistical mechanics framework by directly building a least structured model of the high-dimensional weight space, considering realistic structured data, stochastic gradient descent training, and the computational depth of neural networks. We also consider whether the number of network parameters outnumbers the number of supplied training data, namely, over- or under-parametrization. Our least structured model reveals that the weight spaces of the under-parametrization and over-parameterization cases belong to the same class, in the sense that these weight spaces are well-connected without any hierarchical clustering structure. In contrast, the shallow-network has a broken weight space, characterized by a discontinuous phase transition, thereby clarifying the benefit of depth in deep learning from the angle of high dimensional geometry. Our effective model also reveals that inside a deep network, there exists a liquid-like central part of the architecture in the sense that the weights in this part behave as randomly as possible, providing algorithmic implications. Our data-driven model thus provides a statistical mechanics insight about why deep learning is unreasonably effective in terms of the high-dimensional weight space, and how deep networks are different from shallow ones.
In this paper, we present a learning-based approach to determining acceptance of arguments under several abstract argumentation semantics. More specifically, we propose an argumentation graph neural network (AGNN) that learns a message-passing algorithm to predict the likelihood of an argument being accepted. The experimental results demonstrate that the AGNN can almost perfectly predict the acceptability under different semantics and scales well for larger argumentation frameworks. Furthermore, analysing the behaviour of the message-passing algorithm shows that the AGNN learns to adhere to basic principles of argument semantics as identified in the literature, and can thus be trained to predict extensions under the different semantics - we show how the latter can be done for multi-extension semantics by using AGNNs to guide a basic search. We publish our code at https://github.com/DennisCraandijk/DL-Abstract-Argumentation
The computational prediction algorithm of neural network, or deep learning, has drawn much attention recently in statistics as well as in image recognition and natural language processing. Particularly in statistical application for censored survival data, the loss function used for optimization has been mainly based on the partial likelihood from Cox's model and its variations to utilize existing neural network library such as Keras, which was built upon the open source library of TensorFlow. This paper presents a novel application of the neural network to the quantile regression for survival data with right censoring, which is adjusted by the inverse of the estimated censoring distribution in the check function. The main purpose of this work is to show that the deep learning method could be flexible enough to predict nonlinear patterns more accurately compared to existing quantile regression methods such as traditional linear quantile regression and nonparametric quantile regression with total variation regularization, emphasizing practicality of the method for censored survival data. Simulation studies were performed to generate nonlinear censored survival data and compare the deep learning method with existing quantile regression methods in terms of prediction accuracy. The proposed method is illustrated with two publicly available breast cancer data sets with gene signatures. The method has been built into a package and is freely available at \url{https://github.com/yicjia/DeepQuantreg}.
Modern deep learning methods constitute incredibly powerful tools to tackle a myriad of challenging problems. However, since deep learning methods operate as black boxes, the uncertainty associated with their predictions is often challenging to quantify. Bayesian statistics offer a formalism to understand and quantify the uncertainty associated with deep neural network predictions. This tutorial provides an overview of the relevant literature and a complete toolset to design, implement, train, use and evaluate Bayesian Neural Networks, i.e. Stochastic Artificial Neural Networks trained using Bayesian methods.
Deep learning's recent history has been one of achievement: from triumphing over humans in the game of Go to world-leading performance in image classification, voice recognition, translation, and other tasks. But this progress has come with a voracious appetite for computing power. This article catalogs the extent of this dependency, showing that progress across a wide variety of applications is strongly reliant on increases in computing power. Extrapolating forward this reliance reveals that progress along current lines is rapidly becoming economically, technically, and environmentally unsustainable. Thus, continued progress in these applications will require dramatically more computationally-efficient methods, which will either have to come from changes to deep learning or from moving to other machine learning methods.
Low level image restoration is an integral component of modern artificial intelligence (AI) driven camera pipelines. Most of these frameworks are based on deep neural networks which present a massive computational overhead on resource constrained platform like a mobile phone. In this paper, we propose several lightweight low-level modules which can be used to create a computationally low cost variant of a given baseline model. Recent works for efficient neural networks design have mainly focused on classification. However, low-level image processing falls under the image-to-image' translation genre which requires some additional computational modules not present in classification. This paper seeks to bridge this gap by designing generic efficient modules which can replace essential components used in contemporary deep learning based image restoration networks. We also present and analyse our results highlighting the drawbacks of applying depthwise separable convolutional kernel (a popular method for efficient classification network) for sub-pixel convolution based upsampling (a popular upsampling strategy for low-level vision applications). This shows that concepts from domain of classification cannot always be seamlessly integrated into image-to-image translation tasks. We extensively validate our findings on three popular tasks of image inpainting, denoising and super-resolution. Our results show that proposed networks consistently output visually similar reconstructions compared to full capacity baselines with significant reduction of parameters, memory footprint and execution speeds on contemporary mobile devices.
From an optimization point of view, resource allocation is one of the cornerstones of research for addressing limiting factors commonly arising in applications such as power outages and traffic jams. In this paper, we take a data-driven approach to estimate an optimal nodal restoration sequence for immediate recovery of the infrastructure networks after natural disasters such as earthquakes. We generate data from td-INDP, a high-fidelity simulator of optimal restoration strategies for interdependent networks, and employ deep neural networks to approximate those strategies. Despite the fact that the underlying problem is NP-complete, the restoration sequences obtained by our method are observed to be nearly optimal. In addition, by training multiple models---the so-called estimators---for a variety of resource availability levels, our proposed method balances a trade-off between resource utilization and restoration time. Decision-makers can use our trained models to allocate resources more efficiently after contingencies, and in turn, improve the community resilience. Besides their predictive power, such trained estimators unravel the effect of interdependencies among different nodal functionalities in the restoration strategies. We showcase our methodology by the real-world interdependent infrastructure of Shelby County, TN.
Pathologist-defined labels are the gold standard for histopathological data sets, regardless of well-known limitations in consistency for some tasks. To date, some datasets on mitotic figures are available and were used for development of promising deep learning-based algorithms. In order to assess robustness of those algorithms and reproducibility of their methods it is necessary to test on several independent datasets. The influence of different labeling methods of these available datasets is currently unknown. To tackle this, we present an alternative set of labels for the images of the auxiliary mitosis dataset of the TUPAC16 challenge. Additional to manual mitotic figure screening, we used a novel, algorithm-aided labeling process, that allowed to minimize the risk of missing rare mitotic figures in the images. All potential mitotic figures were independently assessed by two pathologists. The novel, publicly available set of labels contains 1,999 mitotic figures (+28.80%) and additionally includes 10,483 labels of cells with high similarities to mitotic figures (hard examples). We found significant difference comparing F_1 scores between the original label set (0.549) and the new alternative label set (0.735) using a standard deep learning object detection architecture. The models trained on the alternative set showed higher overall confidence values, suggesting a higher overall label consistency. Findings of the present study show that pathologists-defined labels may vary significantly resulting in notable difference in the model performance. Comparison of deep learning-based algorithms between independent datasets with different labeling methods should be done with caution.
The impact of gradient noise on training deep models is widely acknowledged but not well understood. In this context, we study the distribution of gradients during training. We introduce a method, Gradient Clustering, to minimize the variance of average mini-batch gradient with stratified sampling. We prove that the variance of average mini-batch gradient is minimized if the elements are sampled from a weighted clustering in the gradient space. We measure the gradient variance on common deep learning benchmarks and observe that, contrary to common assumptions, gradient variance increases during training, and smaller learning rates coincide with higher variance. In addition, we introduce normalized gradient variance as a statistic that better correlates with the speed of convergence compared to gradient variance.
While deep learning methods continue to improve in predictive accuracy on a wide range of application domains, significant issues remain with other aspects of their performance including their ability to quantify uncertainty and their robustness. Recent advances in approximate Bayesian inference hold significant promise for addressing these concerns, but the computational scalability of these methods can be problematic when applied to large-scale models. In this paper, we describe initial work on the development ofURSABench(the Uncertainty, Robustness, Scalability, and Accu-racy Benchmark), an open-source suite of bench-marking tools for comprehensive assessment of approximate Bayesian inference methods with a focus on deep learning-based classification tasks
We present an algorithm to learn the relevant latent variables of a large-scale discretized physical system and predict its time evolution using thermodynamically-consistent deep neural networks. Our method relies on sparse autoencoders, which reduce the dimensionality of the full order model to a set of sparse latent variables with no prior knowledge of the coded space dimensionality. Then, a second neural network is trained to learn the metriplectic structure of those reduced physical variables and predict its time evolution with a so-called structure-preserving neural network. This data-based integrator is guaranteed to conserve the total energy of the system and the entropy inequality, and can be applied to both conservative and dissipative systems. The integrated paths can then be decoded to the original full-dimensional manifold and be compared to the ground truth solution. This method is tested with two examples applied to fluid and solid mechanics.
Electrocardiogram (ECG) is one of the most convenient and non-invasive tools for monitoring peoples' heart condition, which can use for diagnosing a wide range of heart diseases, including Cardiac Arrhythmia, Acute Coronary Syndrome, et al. However, traditional ECG disease detection models show substantial rates of misdiagnosis due to the limitations of the abilities of extracted features. Recent deep learning methods have shown significant advantages, but they do not provide publicly available services for those who have no training data or computational resources. In this paper, we demonstrate our work on building, training, and serving such out-of-the-box cloud deep learning service for cardiac disease detection from ECG named CardioLearn. The analytic ability of any other ECG recording devices can be enhanced by connecting to the Internet and invoke our open API. As a practical example, we also design a portable smart hardware device along with an interactive mobile program, which can collect ECG and detect potential cardiac diseases anytime and anywhere.
Choosing the optimizer is considered to be among the most crucial design decisions in deep learning, and it is not an easy one. The growing literature now lists hundreds of optimization methods. In the absence of clear theoretical guidance and conclusive empirical evidence, the decision is often made based on anecdotes. In this work, we aim to replace these anecdotes, if not with a conclusive ranking, then at least with evidence-backed heuristics. To do so, we perform an extensive, standardized benchmark of fifteen particularly popular deep learning optimizers while giving a concise overview of the wide range of possible choices. Analyzing more than $50,000$ individual runs, we contribute the following three points: (i) Optimizer performance varies greatly across tasks. (ii) We observe that evaluating multiple optimizers with default parameters works approximately as well as tuning the hyperparameters of a single, fixed optimizer. (iii) While we cannot discern an optimization method clearly dominating across all tested tasks, we identify a significantly reduced subset of specific optimizers and parameter choices that generally lead to competitive results in our experiments: Adam remains a strong contender, with newer methods failing to significantly and consistently outperform it. Our open-sourced results are available as challenging and well-tuned baselines for more meaningful evaluations of novel optimization methods without requiring any further computational efforts.
Vision-based prediction algorithms have a wide range of applications including autonomous driving, surveillance, human-robot interaction, weather prediction. The objective of this paper is to provide an overview of the field in the past five years with a particular focus on deep learning approaches. For this purpose, we categorize these algorithms into video prediction, action prediction, trajectory prediction, body motion prediction, and other prediction applications. For each category, we highlight the common architectures, training methods and types of data used. In addition, we discuss the common evaluation metrics and datasets used for vision-based prediction tasks. A database of all the information presented in this survey including, cross-referenced according to papers, datasets and metrics, can be found online at https://github.com/aras62/vision-based-prediction.
Hybrid beamforming is a promising technique to reduce the complexity and cost of massive multiple-input multiple-output (MIMO) systems while providing high data rate. However, the hybrid precoder design is a challenging task requiring channel state information (CSI) feedback and solving a complex optimization problem. This paper proposes a novel RSSI-based unsupervised deep learning method to design the hybrid beamforming in massive MIMO systems. Furthermore, we propose i) a method to design the synchronization signal (SS) in initial access (IA); and ii) a method to design the codebook for the analog precoder. We also evaluate the system performance through a realistic channel model in various scenarios. We show that the proposed method not only greatly increases the spectral efficiency especially in frequency-division duplex (FDD) communication by using partial CSI feedback, but also has near-optimal sum-rate and outperforms other state-of-the-art full-CSI solutions.
The last decade has witnessed a boom of deep learning research and applications achieving state-of-the-art results in various domains. However, most advances have been established empirically, and their theoretical analysis remains lacking. One major issue is that our current interpretation of neural networks (NNs) as function approximators is too generic to support in-depth analysis. In this paper, we remedy this by proposing an alternative interpretation that identifies NNs as chain graphs (CGs) and feed-forward as an approximate inference procedure. The CG interpretation specifies the nature of each NN component within the rich theoretical framework of probabilistic graphical models, while at the same time remains general enough to cover real-world NNs with arbitrary depth, multi-branching and varied activations, as well as common structures including convolution / recurrent layers, residual block and dropout. We demonstrate with concrete examples that the CG interpretation can provide novel theoretical support and insights for various NN techniques, as well as derive new deep learning approaches such as the concept of partially collapsed feed-forward inference. It is thus a promising framework that deepens our understanding of neural networks and provides a coherent theoretical formulation for future deep learning research.
Consistency of the predictions with respect to the physical forward model is pivotal for reliably solving inverse problems. This consistency is mostly un-controlled in the current end-to-end deep learning methodologies proposed for the Magnetic Resonance Fingerprinting (MRF) problem. To address this, we propose ProxNet, a learned proximal gradient descent framework that directly incorporates the forward acquisition and Bloch dynamic models within a recurrent learning mechanism. The ProxNet adopts a compact neural proximal model for de-aliasing and quantitative inference, that can be flexibly trained on scarce MRF training datasets. Our numerical experiments show that the ProxNet can achieve a superior quantitative inference accuracy, much smaller storage requirement, and a comparable runtime to the recent deep learning MRF baselines, while being much faster than the dictionary matching schemes. Code has been released at https://github.com/edongdongchen/PGD-Net.
Real-world large-scale datasets are heteroskedastic and imbalanced -- labels have varying levels of uncertainty and label distributions are long-tailed. Heteroskedasticity and imbalance challenge deep learning algorithms due to the difficulty of distinguishing among mislabeled, ambiguous, and rare examples. Addressing heteroskedasticity and imbalance simultaneously is under-explored. We propose a data-dependent regularization technique for heteroskedastic datasets that regularizes different regions of the input space differently. Inspired by the theoretical derivation of the optimal regularization strength in a one-dimensional nonparametric classification setting, our approach adaptively regularizes the data points in higher-uncertainty, lower-density regions more heavily. We test our method on several benchmark tasks, including a real-world heteroskedastic and imbalanced dataset, WebVision. Our experiments corroborate our theory and demonstrate a significant improvement over other methods in noise-robust deep learning.