Transfer Learning for Human & AI

Unsupervised Domain Adaptation for semantic segmentation has gained immensepopularity since it can transfer knowledge from simulation to real (Sim2Real) In this work, we present a novel two-phase adaptation scheme . In the first step, we exhaustively distill source domain knowledge using supervised loss functions .…

Heatmap-based methods dominate in the field of human pose estimation by modelling the output distribution through likelihood heatmaps . Residual Log-likelihood Estimation(RLE) is effective, efficient and flexible . Compared to the conventional regression paradigm, regression with RLE bring 12.4 mAPimprovement on MSCOCO without any test-time overhead .…

This paper introduces the sixth Oriental Language Recognition (OLR) 2021Challenge . It intends to improve the performance of language recognitionsystems and speech recognition systems within multilingual scenarios . The dataprofile, four tasks, two baselines, and the evaluation principles are presented .…

SuperCaustics is a real-time,open-source simulation of transparent objects designed for deep learning applications . It uses hardware ray-tracing to support caustics, dispersion, and refraction . The neural network achieved performance comparable to the state-of-the-art on a real world dataset using only 10% of the training data and in a fraction of training time .…

In this work, we consider a single input single output (SISO) system forReconfigurable Intelligent Surface (RIS) assisted mmWave communication . Weconsider Rician channel models over user node to RIS and RIS to Access Point(AP) We obtain closed form expressions for capacity with channel stateinformation (CSI) and without CSI at the transmitter .…

This paper presents HURRA, a system that aims to reduce the time spent by network operators in the process of network troubleshooting . It consists of two modules that are plugged after any anomaly detection algorithm . The main difficulty in live deployment concerns the automated selection of the algorithm and tuning of its hyper-parameters .…

Recent state-of-the-art active learning methods have mostly leveragedGenerative Adversarial Networks (GAN) for sample acquisition . However, GAN isusually known to suffer from instability and sensitivity to hyper-parameters . In contrast to these methods, we propose in this paper a novel active learningframework that we call Maximum Classifier Discrepancy for Active Learning(MCDAL) which takes the prediction discrepancies between multiple classifiers .…

Semantic segmentation on 3D point clouds is an important task for 3D sceneunderstanding . We train a semantic point cloud segmentation network with only asmall portion of points being labeled . We argue that we can better utilize thelimited supervision information as we densely propagate the supervision signalfrom the labeled points to other points within and across the input samples .…

Meta-learning provides a promising way for learning to efficiently learn andachieve great success in many applications . Most meta-learning literature focuses on dealing with tasks from a same domain, making it brittle to generalize to tasks from the other unseen domains .…

Current algorithms share one limitation: They rely on directly visible objects . This is a major drawback compared to human behavior, where indirect visual cues caused by the actual object (e.g.,shadows) are already used intuitively to retrieve information . Humans already process light artifacts caused by oncoming vehicles to assume their future appearance, whereas current objectdetection systems rely on the oncoming vehicle’s direct visibility .…

The combined use of multiple modalities enables accurate pedestrian detection under poor lighting conditions by using the high visibility areas from thesemodalities together . The vital assumption for the combination use is that thereis no or only a weak misalignment between the two modalities .…

Trust is the basis of any distributed, fault-tolerant, or secure system . In systems subject to Byzantine faults, the trust assumption is usually specified through sets of processes that may fail together . Reaching consensus with asymmetric trust in the model of personal Byzantinequorum systems was shown to be impossible, if the trust assumptions of the processes diverge from each other .…

We introduce a runtime-efficient tree hashing algorithm for the identification of isomorphic subtrees . We propose a simple diversity-preservationmechanism with promising results on a collection of symbolic regressionbenchmark problems . The algorithm has two important applications: fast calculation ofpopulation diversity and algebraic simplification of symbolic expression trees .…

Dawar and Wilsenach (ICALP 2020) show an exponential separation between the sizes of symmetric arithmeticcircuits for computing the determinant and the permanent . The symmetryrestriction is that the circuits which take a matrix input are unchanged by a permutation applied simultaneously to the rows and columns of the matrix .…

In this paper we present SurfaceNet, an approach for estimatingspatially-varying bidirectional reflectance distribution function (SVBRDF)material properties from a single image . We pose the problem as an imagetranslation task and propose a novel patch-based generative adversarial network that is able to produce high-quality, high-resolution surface reflectancemaps .…

Organ matching decision is the most critical decision to assign limited viable organs to the most suitable patients . Currently, organ matching decisions were only made by matching scores calculated viascoring models . AWTR improves performance of the current scoring models by using limited actual matching performance in historical data set as well as thecollected covariates from organ donors and patients .…

This letter investigates a downlink multiple input single output (MISO)system based on transmissive reconfigurable metasurface (RMS) transmitter . It proposes an alternating optimization (AO) techniquebased on difference-of-convex (DC) programming and successive convexapproximation (SCA) Simulation results verify that the proposed algorithm canachieve convergence and improve the achievable sum-rate of the system .…

Reinforcement learning (RL) can be used to learn treatment policies and aiddecision making in healthcare . However, given the need for generalization overcomplex state/action spaces, the incorporation of function approximators (e.g.,deep neural networks) requires model selection to reduce overfitting and improve policy performance at deployment .…

Generative adversarial networks (GANs) studies have grown exponentially in the past few years . Their impact has been seen mainly in computer vision with realistic image and video manipulation, especially generation, making significant advancements . GAN applications have diversified across disciplinessuch as time series and sequence generation .…

A paper develops a new shape model that allows synthesizing noveldistance views by optimizing a continuous signed directional distance function . Unlike an SDF, an SDDF measures distance in a given direction . This allows us to form a shape model without 3D shape supervision, using only distancemeasurements, readily available from depth camera or Lidar sensors .…

Graph neural networks (GNNs) have achieved remarkable success as a framework for deep learning on graph-structured data . However, GNNs are fundamentallylimited by their tree-structuring inductive bias . We propose to augment the GNN message-passing operations with information on ego graphs (i.e.,…

In order to define the “shortest path”, a cost function has to bespecified based on theories and heuristics in the application domain . “Ideal” here is defined as guiding the algorithm to plan routes that are most similar to those chosen by humans .…

Recent state-of-the-art active learning methods have mostly leveragedGenerative Adversarial Networks (GAN) for sample acquisition . However, GAN isusually known to suffer from instability and sensitivity to hyper-parameters . In contrast to these methods, we propose in this paper a novel active learningframework that we call Maximum Classifier Discrepancy for Active Learning(MCDAL) which takes the prediction discrepancies between multiple classifiers .…

Deep learning models have gained great popularity in statistical modeling . Theadvantage of deep learning models is that their solutions are difficult tointerpret and explain . We propose a new network architecture that sharessimilar features as generalized linear models, but provides superior predictivepower benefiting from the art of representation learning .…

This paper studies distributed algorithms for (strongly convex) compositeoptimization problems over mesh networks . Instead of focusing on a specific algorithmic design, we propose a black-boxmodel casting distributed algorithms in the form of fixed-point iterates,converging at linear rate . The algorithmic model is coupled with a novel(random) Biased Compression (BC-)rule on the quantizer design, which preserves linear convergence .…

We propose a novel method for solving regression tasks using few-shot or weaksupervision . At the core of our method is the fundamental observation that GANsare incredibly successful at encoding semantic information within their latentspace, even in a completely unsupervised setting .…

Scene text editing (STE) is a challenging task due to a complex intervention between text and style . We propose a novel representational learning-based STE model that employs textual information as well as visual information . Our experiments demonstrate that RewriteNet achieves betterquantitative and qualitative performance than other comparisons .…

Point clouds acquired from scanning devices are often perturbed by noise, which affects downstream tasks such as surface reconstruction and analysis . To denoise a noisy point cloud, we propose to increase the log-likelihood of each point from $p *n$ via gradient ascent .…

Octo-Tiger is a code for modeling three-dimensional self-gravitatingastrophysical fluids . It was particularly designed for the study of dynamicalmass transfer between interacting binary stars . We present scaling results for the newhydro kernels on ORNL’s Summit machine using a Sedov-Taylor blast wave problem .…

All-digital basestation (BS) architectures for millimeter-wave (mmWave)massive multi-user multiple-input multiple- input multiple-output (MU-MIMO) have advantages inspectral efficiency, flexibility, and baseband-processing simplicity overhybrid analog-digital solutions . We demonstrate that adapting the resolution of the analog-to-digitalconverters (ADCs) and spatial equalizer of an all-digital system to the communication scenario enables orders-of-magnitude power savings for realisticmmWave channels .…

Session types enable the static verification of message-passing programs . Asession type specifies a channel’s protocol as sequences of messages . This paper establishes a minimality result but now for the session\pi-calculus, the language in which values are names and for which sessiontypes have been more widely studied .…

In this paper, we connect the tools from cryptology and complexity theory in the domain of Boolean functions with low polynomial degree and high sensitivity . We show that one can implement resilientBoolean functions on a large number of variables with linear size andlogarithmic depth .…

We report gains for both zero-shot and few-shot learning setups, up to 2.7 accuracy points on average, which are even more prominent for low-resource languages (e.g., HaitianCreole) We evaluate our novel latent translation-based model on a series ofmultilingual NLU tasks, including commonsense reasoning, paraphraseidentification, and natural language inference .…

A new method that amends the label distribution of each facial image byleveraging correlations among expressions in the semantic space . By utilizing label distribution learning, a probability distribution is assigned for a facial image to express a compound emotion, which effectivelyimproves the problem of label uncertainties and noises occurred in one-hotlabels .…

Malware detection and analysis are active research subjects in cybersecurity over the last years . The usual detection methods do not necessarily provide tools to interpret the results . The proposed model can determine if a sample ispacked or encrypted with a precision of 85% .…

Privacy is important for all individuals in everyday life . With emergingtechnologies, smartphones with AR, various social networking applications and modes of surveillance, they tend to intrudeprivacy . This study aimed to explore and discover various concerns related toperception of privacy in this era of ubiquitous technologies .…

The method significantly compressesneural network weights to a sparse ternary of [-1,0,1 . It can compress aResNet-18 model from 46 MB to 955KB and a ResNet-50 model from 99 MB to 3.3MB (~30x) The top-1 accuracy on ImageNet drops slightly from 69.7% to65.3% and from 76.15% to 74.47% .…

Image inpainting aims to complete the missing or corrupted regions of images with realistic contents . WaveFill decomposes images by using discrete wavelettransform (DWT) that preserves spatial information naturally . It applies L1reconstruction loss to the decomposed low-frequency bands and adversarial loss .…

Constellation is anetwork that learns relational abstractions of static visual scenes, and generalises these abstractions over sensory particularities . We further show that thisbasis, along with language association, provides a means to imagine sensorycontent in new ways . This work is a first step in the explicit representation of visual relationships and using them for complex cognitive procedures .…

Conditional computation and modular networks have been recently proposed formultitask learning and other problems as a way to decompose problem solving into multiple reusable computational blocks . We propose a new approach forlearning modular networks based on the isometric version of ResNet with allresidual blocks having the same configuration and the same number of parameters .…

Deep generative models have made great progress in synthesizing images with human poses and transferring poses of one person to others . Most existing approaches explicitly leverage the pose information extracted from the source images as a conditional input for the generative networks .…

This paper introduces network flexibility into the chance constrainedeconomic dispatch (CCED) In the proposed model, both power generations andline susceptances become variables to minimize the expected generation cost and guarantee a low probability of constraint violation . We figure out the mechanism of network flexibility against uncertainties from the analytical form of CCED .…

Compact convolutional neural networks (CNNs) have witnessed exceptionalimprovements in performance in recent years . However, they still fail toprovide the same predictive power as CNNs with a large number of parameters . Diverse features present in activation maps indicate presence of unique descriptors necessary to distinguish between objects of differentclasses .…

Understanding emotions that people express during large-scale crises helps inform policy makers and first responders about theemotional states of the population . We present CovidEmo, ~1K tweets labeled withemotions. We examine how well large pre-trained language models generalizeacross domains and crises in the task of perceived emotion prediction .…

This paper addresses the problem of active information gathering formulti-robot systems . The majority of existing information gathering approaches are centralized and, therefore, they cannot be applied to distributed robot teamswhere communication to a central user is not available . In our non-myopic approach, all robots build in parallellocal trees exploring the information space and their corresponding motionspace .…

Compact convolutional neural networks (CNNs) have witnessed exceptionalimprovements in performance in recent years . However, they still fail toprovide the same predictive power as CNNs with a large number of parameters . Diverse features present in activation maps indicate presence of unique descriptors necessary to distinguish between objects of differentclasses .…

Explainability of deep neural networks is one of the most challenging and interesting problems in the field . We show that such an approach can decompose the highly convoluted latent latent spaces of registration pipelines in an orthogonal space with several interesting properties .…

Value iteration is a well-known method of solving Markov Decision Processes . However, the computational cost of value iteration quickly becomesfeasible as the size of the state space increases . In this paper, we propose an intuitive algorithm for solving MDPsthat reduces the cost of updates by dynamically grouping together states with similar cost-to-go values .…

Meta-learning provides a promising way for learning to efficiently learn andachieve great success in many applications . Most meta-learning literature focuses on dealing with tasks from a same domain, making it brittle to generalize to tasks from the other unseen domains .…

Federated Learning (FL) is known to perform Machine Learning tasks in adistributed manner . A challenge that exists in FL is the communication costs, as FLtakes place in a distributed environment where devices connected over thenetwork have to constantly share their updates this can create a communication bottleneck .…