With the recent success of representation learning methods, which includes deep learning as a special case, there has been considerable interest in developing representation learning techniques that can incorporate known physical constraints into the learned representation. As one example, in many applications that involve a signal propagating through physical media (e.g., optics, acoustics, fluid dynamics, etc), it is known that the dynamics of the signal must satisfy constraints imposed by the wave equation. Here we propose a matrix factorization technique that decomposes such signals into a sum of components, where each component is regularized to ensure that it satisfies wave equation constraints. Although our proposed formulation is non-convex, we prove that our model can be efficiently solved to global optimality in polynomial time. We demonstrate the benefits of our work by applications in structural health monitoring, where prior work has attempted to solve this problem using sparse dictionary learning approaches that do not come with any theoretical guarantees regarding convergence to global optimality and employ heuristics to capture desired physical constraints.

With the increasing reliance of users on smart devices, bringing essential computation at the edge has become a crucial requirement for any type of business. Many such computations utilize Convolution Neural Networks (CNNs) to perform AI tasks, having high resource and computation requirements, that are infeasible for edge devices. Splitting the CNN architecture to perform part of the computation on edge and remaining on the cloud is an area of research that has seen increasing interest in the field. In this paper, we assert that running CNNs between an edge device and the cloud is synonymous to solving a resource-constrained optimization problem that minimizes the latency and maximizes resource utilization at the edge. We formulate a multi-objective optimization problem and propose the LMOS algorithm to achieve a Pareto efficient solution. Experiments done on real-world edge devices show that, LMOS ensures feasible execution of different CNN models at the edge and also improves upon existing state-of-the-art approaches.

Current datasets to train social behaviors are usually borrowed from surveillance applications that capture visual data from a bird's-eye perspective. This leaves aside precious relationships and visual cues that could be captured through a first-person view of a scene. In this work, we propose a strategy to exploit the power of current game engines, such as Unity, to transform pre-existing bird's-eye view datasets into a first-person view, in particular, a depth view. Using this strategy, we are able to generate large volumes of synthetic data that can be used to pre-train a social navigation model. To test our ideas, we present DeepSocNav, a deep learning based model that takes advantage of the proposed approach to generate synthetic data. Furthermore, DeepSocNav includes a self-supervised strategy that is included as an auxiliary task. This consists of predicting the next depth frame that the agent will face. Our experiments show the benefits of the proposed model that is able to outperform relevant baselines in terms of social navigation scores.

Magnetic resonance imaging (MRI) is a medical imaging technique used to capture volumetric image sequences of internal soft tissues, such as cardiac muscles. In comparison to X-Ray imaging (XR) and Computer Tomography (CT), MRI provides images with improved structural details via finer spatial resolutions. Cardiac MRI (CMR) focuses on the heart, allowing trained cardiologists to measure heart parameters, for example the mass of the cardiac muscle (myocardium mass), the volumes of blood cavities (atrial and ventricular volumes) and the amount of blood pumped per heartbeat (ejection fraction) [Peng et al., 2016]. Those parameters are used to assess how healthy is the heart, by recognizing early conditions and signs before the onset of infarcts and other complications. Due to the size and complexity of CMR sequences, complex techniques are required to produce detailed analyses; one of these techniques is deep learning (DL). Many of the tasks and goals related to the cardiac functional analysis - for example segmentation of structures [Bernard et al., 2018], estimation of heart parameters [Xue et al., 2018], and detection of diseases [Khened et al., 2017] - have benefited from DL methods. For even better results, research in DL has pointed out that models based on convolutional neural networks (CNN) have had a higher efficacy when provided with regions-of-interest (ROI) either explicitly or implicitly [Xue et al., 2018]. The detection of ROIs, usually named ROI proposal, is a preprocessing step whose goal is to identify the most prominent regions of an image (frame) for discovering clinically-relevant artifacts. The explicit ROI proposal approaches usually follow a combination of methods, for example: (a) pipelining a segmentation and a regression network; (b) preprocessing the input with a region proposal algorithm [He et al., 2015] or with a CNN [Wu et al., 2020]; or (c) by using manual cropping [Xue et al., 2017].

Collecting large-scale annotated satellite imagery datasets is essential for deep-learning-based global building change surveillance. In particular, the scroll imaging mode of optical satellites enables larger observation ranges and shorter revisit periods, facilitating efficient global surveillance. However, the images in recent satellite change detection datasets are mainly captured at near-nadir viewing angles. In this paper, we introduce S2Looking, a building change detection dataset that contains large-scale side-looking satellite images captured at varying off-nadir angles. Our S2Looking dataset consists of 5000 registered bitemporal image pairs (size of 1024*1024, 0.5 ~ 0.8 m/pixel) of rural areas throughout the world and more than 65,920 annotated change instances. We provide two label maps to separately indicate the newly built and demolished building regions for each sample in the dataset. We establish a benchmark task based on this dataset, i.e., identifying the pixel-level building changes in the bi-temporal images. We test several state-of-the-art methods on both the S2Looking dataset and the (near-nadir) LEVIR-CD+ dataset. The experimental results show that recent change detection methods exhibit much poorer performance on the S2Looking than on LEVIR-CD+. The proposed S2Looking dataset presents three main challenges: 1) large viewing angle changes, 2) large illumination variances and 3) various complex scene characteristics encountered in rural areas. Our proposed dataset may promote the development of algorithms for satellite image change detection and registration under conditions of large off-nadir angles. The dataset is available at https://github.com/AnonymousForACMMM/.

In this work, we modify and apply self-supervision techniques to the domain of medical health insurance claims. We model patients' healthcare claims history analogous to free-text narratives, and introduce pre-trained `prior knowledge', later utilized for patient outcome predictions on a challenging task: predicting Covid-19 hospitalization, given a patient's pre-Covid-19 insurance claims history. Results suggest that pre-training on insurance claims not only produces better prediction performance, but, more importantly, improves the model's `clinical trustworthiness' and model stability/reliability.

With the advancements in computer technology, there is a rapid development of intelligent systems to understand the complex relationships in data to make predictions and classifications. Artificail Intelligence based framework is rapidly revolutionizing the healthcare industry. These intelligent systems are built with machine learning and deep learning based robust models for early diagnosis of diseases and demonstrates a promising supplementary diagnostic method for frontline clinical doctors and surgeons. Machine Learning and Deep Learning based systems can streamline and simplify the steps involved in diagnosis of diseases from clinical and image-based data, thus providing significant clinician support and workflow optimization. They mimic human cognition and are even capable of diagnosing diseases that cannot be diagnosed with human intelligence. This paper focuses on the survey of machine learning and deep learning applications in across 16 medical specialties, namely Dental medicine, Haematology, Surgery, Cardiology, Pulmonology, Orthopedics, Radiology, Oncology, General medicine, Psychiatry, Endocrinology, Neurology, Dermatology, Hepatology, Nephrology, Ophthalmology, and Drug discovery. In this paper along with the survey, we discuss the advancements of medical practices with these systems and also the impact of these systems on medical professionals.

A fundamental challenge for any intelligent system is prediction: given some inputs $X_1,..,X_\tau$ can you predict outcomes $Y_1,.., Y_\tau$. The KL divergence $\mathbf{d}_{\mathrm{KL}}$ provides a natural measure of prediction quality, but the majority of deep learning research looks only at the marginal predictions per input $X_t$. In this technical report we propose a scoring rule $\mathbf{d}_{\mathrm{KL}}^\tau$, parameterized by $\tau \in \mathcal{N}$ that evaluates the joint predictions at $\tau$ inputs simultaneously. We show that the commonly-used $\tau=1$ can be insufficient to drive good decisions in many settings of interest. We also show that, as $\tau$ grows, performing well according to $\mathbf{d}_{\mathrm{KL}}^\tau$ recovers universal guarantees for any possible decision. Finally, we provide problem-dependent guidance on the scale of $\tau$ for which our score provides sufficient guarantees for good performance.

Recent deep-learning models have achieved impressive prediction performance, but often sacrifice interpretability and computational efficiency. Interpretability is crucial in many disciplines, such as science and medicine, where models must be carefully vetted or where interpretation is the goal itself. Moreover, interpretable models are concise and often yield computational efficiency. Here, we propose adaptive wavelet distillation (AWD), a method which aims to distill information from a trained neural network into a wavelet transform. Specifically, AWD penalizes feature attributions of a neural network in the wavelet domain to learn an effective multi-resolution wavelet transform. The resulting model is highly predictive, concise, computationally efficient, and has properties (such as a multi-scale structure) which make it easy to interpret. In close collaboration with domain experts, we showcase how AWD addresses challenges in two real-world settings: cosmological parameter inference and molecular-partner prediction. In both cases, AWD yields a scientifically interpretable and concise model which gives predictive performance better than state-of-the-art neural networks. Moreover, AWD identifies predictive features that are scientifically meaningful in the context of respective domains. All code and models are released in a full-fledged package available on Github (https://github.com/Yu-Group/adaptive-wavelets).

In this work we explore the limiting dynamics of deep neural networks trained with stochastic gradient descent (SGD). We find empirically that long after performance has converged, networks continue to move through parameter space by a process of anomalous diffusion in which distance travelled grows as a power law in the number of gradient updates with a nontrivial exponent. We reveal an intricate interaction between the hyperparameters of optimization, the structure in the gradient noise, and the Hessian matrix at the end of training that explains this anomalous diffusion. To build this understanding, we first derive a continuous-time model for SGD with finite learning rates and batch sizes as an underdamped Langevin equation. We study this equation in the setting of linear regression, where we can derive exact, analytic expressions for the phase space dynamics of the parameters and their instantaneous velocities from initialization to stationarity. Using the Fokker-Planck equation, we show that the key ingredient driving these dynamics is not the original training loss, but rather the combination of a modified loss, which implicitly regularizes the velocity, and probability currents, which cause oscillations in phase space. We identify qualitative and quantitative predictions of this theory in the dynamics of a ResNet-18 model trained on ImageNet. Through the lens of statistical physics, we uncover a mechanistic origin for the anomalous limiting dynamics of deep neural networks trained with SGD.