This work introduces the Topological CNN (TCNN), which encompasses several topologically defined convolutional methods. Manifolds with important relationships to the natural image space are used to parameterize image filters which are used as convolutional weights in a TCNN. These manifolds also parameterize slices in layers of a TCNN across which the weights are localized. We show evidence that TCNNs learn faster, on less data, with fewer learned parameters, and with greater generalizability and interpretability than conventional CNNs. We introduce and explore TCNN layers for both image and video data. We propose extensions to 3D images and 3D video.
Deep learning has become the gold standard for image processing over the past decade. Simultaneously, we have seen growing interest in orbital activities such as satellite servicing and debris removal that depend on proximity operations between spacecraft. However, two key challenges currently pose a major barrier to the use of deep learning for vision-based on-orbit proximity operations. Firstly, efficient implementation of these techniques relies on an effective system for model development that streamlines data curation, training, and evaluation. Secondly, a scarcity of labeled training data (images of a target spacecraft) hinders creation of robust deep learning models. This paper presents an open-source deep learning pipeline, developed specifically for on-orbit visual navigation applications, that addresses these challenges. The core of our work consists of two custom software tools built on top of a cloud architecture that interconnects all stages of the model development process. The first tool leverages Blender, an open-source 3D graphics toolset, to generate labeled synthetic training data with configurable model poses (positions and orientations), lighting conditions, backgrounds, and commonly observed in-space image aberrations. The second tool is a plugin-based framework for effective dataset curation and model training; it provides common functionality like metadata generation and remote storage access to all projects while giving complete independence to project-specific code. Time-consuming, graphics-intensive processes such as synthetic image generation and model training run on cloud-based computational resources which scale to any scope and budget and allow development of even the largest datasets and models from any machine. The presented system has been used in the Texas Spacecraft Laboratory with marked benefits in development speed and quality.
We study the performance of Long Short-Term Memory networks for keystroke biometric authentication at large scale in free-text scenarios. For this we explore the performance of Long Short-Term Memory (LSTMs) networks trained with a moderate number of keystrokes per identity and evaluated under different scenarios including: i) three learning approaches depending on the loss function (softmax, contrastive, and triplet loss); ii) different number of training samples and lengths of keystroke sequences; iii) four databases based on two device types (physical vs touchscreen keyboard); and iv) comparison with existing approaches based on both traditional statistical methods and deep learning architectures. Our approach called TypeNet achieves state-of-the-art keystroke biometric authentication performance with an Equal Error Rate of 2.2% and 9.2% for physical and touchscreen keyboards, respectively, significantly outperforming previous approaches. Our experiments demonstrate a moderate increase in error with up to 100,000 subjects, demonstrating the potential of TypeNet to operate at an Internet scale. To the best of our knowledge, the databases used in this work are the largest existing free-text keystroke databases available for research with more than 136 million keystrokes from 168,000 subjects in physical keyboards, and 60,000 subjects with more than 63 million keystrokes acquired on mobile touchscreens.
This paper addresses the growing need to process non-Euclidean data, by introducing a geometric deep learning (GDL) framework for building universal feedforward-type models compatible with differentiable manifold geometries. We show that our GDL models can approximate any continuous target function uniformly on compact sets of a controlled maximum diameter. We obtain curvature-dependent lower-bounds on this maximum diameter and upper-bounds on the depth of our approximating GDL models. Conversely, we find that there is always a continuous function between any two non-degenerate compact manifolds that any "locally-defined" GDL model cannot uniformly approximate. Our last main result identifies data-dependent conditions guaranteeing that the GDL model implementing our approximation breaks "the curse of dimensionality." We find that any "real-world" (i.e. finite) dataset always satisfies our condition and, conversely, any dataset satisfies our requirement if the target function is smooth. As applications, we confirm the universal approximation capabilities of the following GDL models: Ganea et al. (2018)'s hyperbolic feedforward networks, the architecture implementing Krishnan et al. (2015)'s deep Kalman-Filter, and deep softmax classifiers. We build universal extensions/variants of: the SPD-matrix regressor of Meyer et al. (2011), and Fletcher (2003)'s Procrustean regressor. In the Euclidean setting, our results imply a quantitative version of Kidger and Lyons (2020)'s approximation theorem and a data-dependent version of Yarotsky and Zhevnerchuk (2019)'s uncursed approximation rates.
We present an introduction to model-based machine learning for communication systems. We begin by reviewing existing strategies for combining model-based algorithms and machine learning from a high level perspective, and compare them to the conventional deep learning approach which utilizes established deep neural network (DNN) architectures trained in an end-to-end manner. Then, we focus on symbol detection, which is one of the fundamental tasks of communication receivers. We show how the different strategies of conventional deep architectures, deep unfolding, and DNN-aided hybrid algorithms, can be applied to this problem. The last two approaches constitute a middle ground between purely model-based and solely DNN-based receivers. By focusing on this specific task, we highlight the advantages and drawbacks of each strategy, and present guidelines to facilitate the design of future model-based deep learning systems for communications.
Deep-learning (DL) has emerged as a powerful machine-learning technique for several classic problems encountered in generic wireless communications. Specifically, random Fourier Features (RFF) based deep-learning has emerged as an attractive solution for several machine-learning problems; yet there is a lacuna of rigorous results to justify the viability of RFF based DL-algorithms in general. To address this gap, we attempt to analytically quantify the viability of RFF based DL. Precisely, in this paper, analytical proofs are presented demonstrating that RFF based DL architectures have lower approximation-error and probability of misclassification as compared to classical DL architectures. In addition, a new distribution-dependent RFF is proposed to facilitate DL architectures with low training-complexity. Through computer simulations, the practical application of the presented analytical results and the proposed distribution-dependent RFF, are depicted for various machine-learning problems encountered in next-generation communication systems such as: a) line of sight (LOS)/non-line of sight (NLOS) classification, and b) message-passing based detection of low-density parity check codes (LDPC) codes over nonlinear visible light communication (VLC) channels. Especially in the low training-data regime, the presented simulations show that significant performance gains are achieved when utilizing RFF maps of observations. Lastly, in all the presented simulations, it is observed that the proposed distribution-dependent RFFs significantly outperform RFFs, which make them useful for potential machine-learning/DL based applications in the context of next-generation communication systems.
Huge overhead of beam training imposes a significant challenge in millimeter-wave (mmWave) wireless communications. To address this issue, in this paper, we propose a wide beam based training approach to calibrate the narrow beam direction according to the channel power leakage. To handle the complex nonlinear properties of the channel power leakage, deep learning is utilized to predict the optimal narrow beam directly. Specifically, three deep learning assisted calibrated beam training schemes are proposed. The first scheme adopts convolution neural network to implement the prediction based on the instantaneous received signals of wide beam training. We also perform the additional narrow beam training based on the predicted probabilities for further beam direction calibrations. However, the first scheme only depends on one wide beam training, which lacks the robustness to noise. To tackle this problem, the second scheme adopts long-short term memory (LSTM) network for tracking the movement of users and calibrating the beam direction according to the received signals of prior beam training, in order to enhance the robustness to noise. To further reduce the overhead of wide beam training, our third scheme, an adaptive beam training strategy, selects partial wide beams to be trained based on the prior received signals. Two criteria, namely, optimal neighboring criterion and maximum probability criterion, are designed for the selection. Furthermore, to handle mobile scenarios, auxiliary LSTM is introduced to calibrate the directions of the selected wide beams more precisely. Simulation results demonstrate that our proposed schemes achieve significantly higher beamforming gain with smaller beam training overhead compared with the conventional and existing deep-learning based counterparts.
Medical brain image analysis is a necessary step in Computer Assisted /Aided Diagnosis (CAD) systems. Advancements in both hardware and software in the past few years have led to improved segmentation and classification of various diseases. In the present work, we review the published literature on systems and algorithms that allow for classification, identification, and detection of White Matter Hyperintensities (WMHs) of brain MRI images specifically in cases of ischemic stroke and demyelinating diseases. For the selection criteria, we used the bibliometric networks. Out of a total of 140 documents we selected 38 articles that deal with the main objectives of this study. Based on the analysis and discussion of the revised documents, there is constant growth in the research and proposal of new models of deep learning to achieve the highest accuracy and reliability of the segmentation of ischemic and demyelinating lesions. Models with indicators (Dice Score, DSC: 0.99) were found, however with little practical application due to the uses of small datasets and lack of reproducibility. Therefore, the main conclusion is to establish multidisciplinary research groups to overcome the gap between CAD developments and their complete utilization in the clinical environment.
The Marathi language is one of the prominent languages used in India. It is predominantly spoken by the people of Maharashtra. Over the past decade, the usage of language on online platforms has tremendously increased. However, research on Natural Language Processing (NLP) approaches for Marathi text has not received much attention. Marathi is a morphologically rich language and uses a variant of the Devanagari script in the written form. This works aims to provide a comprehensive overview of available resources and models for Marathi text classification. We evaluate CNN, LSTM, ULMFiT, and BERT based models on two publicly available Marathi text classification datasets and present a comparative analysis. The pre-trained Marathi fast text word embeddings by Facebook and IndicNLP are used in conjunction with word-based models. We show that basic single layer models based on CNN and LSTM coupled with FastText embeddings perform on par with the BERT based models on the available datasets. We hope our paper aids focused research and experiments in the area of Marathi NLP.
The volume of "free" data on the internet has been key to the current success of deep learning. However, it also raises privacy concerns about the unauthorized exploitation of personal data for training commercial models. It is thus crucial to develop methods to prevent unauthorized data exploitation. This paper raises the question: \emph{can data be made unlearnable for deep learning models?} We present a type of \emph{error-minimizing} noise that can indeed make training examples unlearnable. Error-minimizing noise is intentionally generated to reduce the error of one or more of the training example(s) close to zero, which can trick the model into believing there is "nothing" to learn from these example(s). The noise is restricted to be imperceptible to human eyes, and thus does not affect normal data utility. We empirically verify the effectiveness of error-minimizing noise in both sample-wise and class-wise forms. We also demonstrate its flexibility under extensive experimental settings and practicability in a case study of face recognition. Our work establishes an important first step towards making personal data unexploitable to deep learning models.
We propose reproducing activation functions (RAFs) to improve deep learning accuracy for various applications ranging from computer vision to scientific computing. The idea is to employ several basic functions and their learnable linear combination to construct neuron-wise data-driven activation functions for each neuron. Armed with RAFs, neural networks (NNs) can reproduce traditional approximation tools and, therefore, approximate target functions with a smaller number of parameters than traditional NNs. In NN training, RAFs can generate neural tangent kernels (NTKs) with a better condition number than traditional activation functions lessening the spectral bias of deep learning. As demonstrated by extensive numerical tests, the proposed RAFs can facilitate the convergence of deep learning optimization for a solution with higher accuracy than existing deep learning solvers for audio/image/video reconstruction, PDEs, and eigenvalue problems. With RAFs, the errors of audio/video reconstruction, PDEs, and eigenvalue problems are decreased by over 14%, 73%, 99%, respectively, compared with baseline, while the performance of image reconstruction increases by 58%.
We demonstrate high fidelity enhancement of planetary digital elevation models (DEMs) using optical images and deep learning with convolutional neural networks. Enhancement can be applied recursively to the limit of available optical data, representing a 90x resolution improvement in global Mars DEMs. Deep learning-based photoclinometry robustly recovers features obscured by non-ideal lighting conditions. Method can be automated at global scale. Analysis shows enhanced DEM slope errors are comparable with high resolution maps using conventional, labor intensive methods.
Standard machine learning approaches require centralizing the users' data in one computer or a shared database, which raises data privacy and confidentiality concerns. Therefore, limiting central access is important, especially in healthcare settings, where data regulations are strict. A potential approach to tackling this is Federated Learning (FL), which enables multiple parties to collaboratively learn a shared prediction model by using parameters of locally trained models while keeping raw training data locally. In the context of AI-assisted pain-monitoring, we wish to enable confidentiality-preserving and unobtrusive pain estimation for long-term pain-monitoring and reduce the burden on the nursing staff who perform frequent routine check-ups. To this end, we propose a novel Personalized Federated Deep Learning (PFDL) approach for pain estimation from face images. PFDL performs collaborative training of a deep model, implemented using a lightweight CNN architecture, across different clients (i.e., subjects) without sharing their face images. Instead of sharing all parameters of the model, as in standard FL, PFDL retains the last layer locally (used to personalize the pain estimates). This (i) adds another layer of data confidentiality, making it difficult for an adversary to infer pain levels of the target subject, while (ii) personalizing the pain estimation to each subject through local parameter tuning. We show using a publicly available dataset of face videos of pain (UNBC-McMaster Shoulder Pain Database), that PFDL performs comparably or better than the standard centralized and FL algorithms, while further enhancing data privacy. This, has the potential to improve traditional pain monitoring by making it more secure, computationally efficient, and scalable to a large number of individuals (e.g., for in-home pain monitoring), providing timely and unobtrusive pain measurement.
In recent years, face biometric security systems are rapidly increasing, therefore, the presentation attack detection (PAD) has received significant attention from research communities and has become a major field of research. Researchers have tackled the problem with various methods, from exploiting conventional texture feature extraction such as LBP, BSIF, and LPQ to using deep neural networks with different architectures. Despite the results each of these techniques has achieved for a certain attack scenario or dataset, most of them still failed to generalized the problem for unseen conditions, as the efficiency of each is limited to certain type of presentation attacks and instruments (PAI). In this paper, instead of completely extracting hand-crafted texture features or relying only on deep neural networks, we address the problem via fusing both wide and deep features in a unified neural architecture. The main idea is to take advantage of the strength of both methods to derive well-generalized solution for the problem. We also evaluated the effectiveness of our method by comparing the results with each of the mentioned techniques separately. The procedure is done on different spoofing datasets such as ROSE-Youtu, SiW and NUAA Imposter datasets. In particular, we simultanously learn a low dimensional latent space empowered with data-driven features learnt via Convolutional Neural Network designes for spoofing detection task (i.e., deep channel) as well as leverages spoofing detection feature already popular for spoofing in frequency and temporal dimensions ( i.e., via wide channel).
Objective: To assess the ability of imaging-based deep learning to predict radiographic patellofemoral osteoarthritis (PFOA) from knee lateral view radiographs. Design: Knee lateral view radiographs were extracted from The Multicenter Osteoarthritis Study (MOST) (n = 18,436 knees). Patellar region-of-interest (ROI) was first automatically detected, and subsequently, end-to-end deep convolutional neural networks (CNNs) were trained and validated to detect the status of patellofemoral OA. Patellar ROI was detected using deep-learning-based object detection method. Manual PFOA status assessment provided in the MOST dataset was used as a classification outcome for the CNNs. Performance of prediction models was assessed using the area under the receiver operating characteristic curve (ROC AUC) and the average precision (AP) obtained from the precision-recall (PR) curve in the stratified 5-fold cross validation setting. Results: Of the 18,436 knees, 3,425 (19%) had PFOA. AUC and AP for the reference model including age, sex, body mass index (BMI), the total Western Ontario and McMaster Universities Arthritis Index (WOMAC) score, and tibiofemoral Kellgren-Lawrence (KL) grade to predict PFOA were 0.806 and 0.478, respectively. The CNN model that used only image data significantly improved the prediction of PFOA status (ROC AUC= 0.958, AP= 0.862). Conclusion: We present the first machine learning based automatic PFOA detection method. Furthermore, our deep learning based model trained on patella region from knee lateral view radiographs performs better at predicting PFOA than models based on patient characteristics and clinical assessments.
Chest X-ray becomes one of the most common medical diagnoses due to its noninvasiveness. The number of chest X-ray images has skyrocketed, but reading chest X-rays still have been manually performed by radiologists, which creates huge burnouts and delays. Traditionally, radiomics, as a subfield of radiology that can extract a large number of quantitative features from medical images, demonstrates its potential to facilitate medical imaging diagnosis before the deep learning era. With the rise of deep learning, the explainability of deep neural networks on chest X-ray diagnosis remains opaque. In this study, we proposed a novel framework that leverages radiomics features and contrastive learning to detect pneumonia in chest X-ray. Experiments on the RSNA Pneumonia Detection Challenge dataset show that our model achieves superior results to several state-of-the-art models (> 10% in F1-score) and increases the model's interpretability.
Deep learning has shown its power in many applications, including object detection in images, natural-language understanding, and speech recognition. To make it more accessible to end users, many deep learning models are now embedded in mobile apps. Compared to offloading deep learning from smartphones to the cloud, performing machine learning on-device can help improve latency, connectivity, and power consumption. However, most deep learning models within Android apps can easily be obtained via mature reverse engineering, while the models' exposure may invite adversarial attacks. In this study, we propose a simple but effective approach to hacking deep learning models using adversarial attacks by identifying highly similar pre-trained models from TensorFlow Hub. All 10 real-world Android apps in the experiment are successfully attacked by our approach. Apart from the feasibility of the model attack, we also carry out an empirical study that investigates the characteristics of deep learning models used by hundreds of Android apps on Google Play. The results show that many of them are similar to each other and widely use fine-tuning techniques to pre-trained models on the Internet.
A method for active learning of hyperspectral images (HSI) is proposed, which combines deep learning with diffusion processes on graphs. A deep variational autoencoder extracts smoothed, denoised features from a high-dimensional HSI, which are then used to make labeling queries based on graph diffusion processes. The proposed method combines the robust representations of deep learning with the mathematical tractability of diffusion geometry, and leads to strong performance on real HSI.
Skin cancer is the most common cancer worldwide, with melanoma being the deadliest form. Dermoscopy is a skin imaging modality that has shown an improvement in the diagnosis of skin cancer compared to visual examination without support. We evaluate the current state of the art in the classification of dermoscopic images based on the ISIC-2019 Challenge for the classification of skin lesions and current literature. Various deep neural network architectures pre-trained on the ImageNet data set are adapted to a combined training data set comprised of publicly available dermoscopic and clinical images of skin lesions using transfer learning and model fine-tuning. The performance and applicability of these models for the detection of eight classes of skin lesions are examined. Real-time data augmentation, which uses random rotation, translation, shear, and zoom within specified bounds is used to increase the number of available training samples. Model predictions are multiplied by inverse class frequencies and normalized to better approximate actual probability distributions. Overall prediction accuracy is further increased by using the arithmetic mean of the predictions of several independently trained models. The best single model has been published as a web service.
Face super-resolution (FSR), also known as face hallucination, which is aimed at enhancing the resolution of low-resolution (LR) face images to generate high-resolution (HR) face images, is a domain-specific image super-resolution problem. Recently, FSR has received considerable attention and witnessed dazzling advances with the development of deep learning techniques. To date, few summaries of the studies on the deep learning-based FSR are available. In this survey, we present a comprehensive review of deep learning-based FSR methods in a systematic manner. First, we summarize the problem formulation of FSR and introduce popular assessment metrics and loss functions. Second, we elaborate on the facial characteristics and popular datasets used in FSR. Third, we roughly categorize existing methods according to the utilization of facial characteristics. In each category, we start with a general description of design principles, then present an overview of representative approaches, and then discuss the pros and cons among them. Fourth, we evaluate the performance of some state-of-the-art methods. Fifth, joint FSR and other tasks, and FSR-related applications are roughly introduced. Finally, we envision the prospects of further technological advancement in this field. A curated list of papers and resources to face super-resolution are available at \url{https://github.com/junjun-jiang/Face-Hallucination-Benchmark}