Adversarial attack is aimed at fooling the target classifier with perturbation . CAP-GAN takes account of the idea of pixel-level and feature-level consistency to achieve reasonablepurification under cycle-consistent learning . We utilize theguided attention module and knowledge distillation to convey meaningfulinformation to the purification model .…

## Learning Intra Batch Connections for Deep Metric Learning

The goal of metric learning is to learn a function that maps samples to alower-dimensional space where similar samples lie closer than dissimilar ones . We propose an approach based on message passing networks that takes into account all therelations in a mini-batch .…

## Scan Specific MRI Reconstruction using Zero Shot Physics Guided Deep Learning

Physics-guided deep learning (PG-DL) has emerged as a powerful tool foraccelerated MRI reconstruction . The main challenge for developing scan-specific PG-DL methods is the large number of parameters, making it prone to over-fitting . The proposed approach splits availablemeasurements for each scan into three disjoint sets .…

## Detection and severity classification of COVID 19 in CT images using deep learning

A cascaded system is proposed to segment the lung,detect, localize, and quantify COVID-19 infections from computed tomography(CT) images . The system classifies the severity of the disease asmild, moderate, severe, or critical based on the percentage of infected lungs . The achievedperformance is significantly superior to previous methods for COVI-19 lesionlocalization.…

## A Gated Fusion Network for Dynamic Saliency Prediction

Gated Fusion Network for dynamic saliency(GFSalNet) is the first deep saliency model capable of making predictions in adynamic way via gated fusion mechanism . The model also exploitsspatial and channel-wise attention within a multi-scale architecture that further allows for highly accurate predictions .…

## Translational Equivariance in Kernelizable Attention

Transformer architectures have show remarkable success, they are boundto the computation of all pairwise interactions of input elements and thussuffer from limited scalability . Recent work has been successful by avoidingthe computation of the complete attention matrix, yet leads to problems downthe line .…

## RMS Net Regression and Masking for Soccer Event Spotting

The proposed action spotting task consists in finding the exacttimestamp in which an event occurs . This task fits particularly well for soccervideos, where events correspond to salient actions strictly defined by soccer rules . We enrich our model with two training strategies: the first for data balancing and uniform sampling, the second for maskingambiguous frames and keeping the most discriminative visual cues .…

## Learning image quality assessment by reinforcing task amenable data selection

In this paper, we consider a type of image quality assessment as atask-specific measurement, which can be used to select images that are moreamenable to a given target task, such as image classification or segmentation . We show that the controller-predicted image quality can besignificantly different from the task-specific image quality labels that aremanually defined by humans .…

## Colored Kimia Path24 Dataset Configurations and Benchmarks with Deep Embeddings

The Kimia Path24 dataset has been introduced as a classification andretrieval dataset for digital pathology . Although it provides multi-class data, the color information has been neglected in the process of extracting patches . To address this drawback, we introduce the color version of KimiaPath24 by recreating sample patches from all 24 scans .…

## Plug and Play external and internal priors for image restoration

Plug-and-Play (PnP) methods for image restoration have obtained very good results and popularity . We propose a new PnP scheme combining externaland internal priors . We prove effectiveness of the proposed method in restoring noisy noisy images, both in simulated and real medical settings.…

## Spatio temporal Graph RNN for Point Cloud Prediction

In this paper, we propose an end-to-end learning network aim at predicting future PC frames, based on point-based RNN network . As main novelty, an initiallayer learns topological information of point clouds as geometric features and then uses the learned features to form representative spatio-temporalneighborhoods .…

## Holographic Cell Stiffness Mapping Using Acoustic Stimulation

Cell stiffness is one of the fundamental mechanical properties of the cell and is greatly affected by the intracellular tensional forces, cytoskeletalprestress, and cytoskeleton structure . Accurate assessment of stiffness distribution is essential due to the critical role of single cell mechanobiology in the regulation of many vital processes such as proliferation, adhesion, migration, and motility .…

## OmniDet Surround View Cameras based Multi task Visual Perception Network for Autonomous Driving

Surround View fisheye cameras are commonly deployed in automated driving for360\deg{} near-field sensing around the vehicle . This work presents amulti-task visual perception network . It consists of six primarytasks necessary for an autonomous driving system: depth estimation, visualodometry, semantic segmentation, motion segmentation and object detection, and lens soiling detection .…

## FAT Learning Low Bitwidth Parametric Representation via Frequency Aware Transformation

Frequency-Aware Transformation (FAT) learns to transform network weights in the frequency domain before quantization . When training ResNet-18 andMobileNet-V2 in 4 bits, FAT plus a simple rounding operation already achieves70.5% and 69.2% top-1 accuracy on ImageNet without bells and whistles .…

## Generation for adaption a Gan based approach for 3D Domain Adaption inPoint Cloud

Recent deep networks have achieved good performance on a variety of 3d points classification tasks . However, these models often face challenges in “wildtasks” Unsupervised domain adaptation (UDA) seeks to overcome such a problem without target domain labels . Instead of aligning features betweensource data and target data, we propose a method that use a Generativeadversarial network to generate synthetic data from the source domain so that the output is close to the target domain .…

## Maximizing Joint Entropy for Batch Mode Active Learning of Perceptual Metrics

Active metric learning is the problem of incrementally selecting batches of training data (typically, ordered triplets) to annotate, in order toprogressively improve a learned model of a metric over some input domain asrapidly as possible . Standard approaches, which independently select eachtriplet in a batch, are susceptible to highly correlated batches with many redundant triplets and hence low overall utility .…

## QuickBrowser A Unified Model to Detect and Read Simple Object in Real time

There are many real-life use cases such as barcode scanning or billboardreading where people need to detect objects and read the object contents . This work aims to solve this detect-and-read problem in alightweight way by integrating multi-digit recognition into a one-stage objectdetection model .…

## Capturing Detailed Deformations of Moving Human Bodies

New method to capture detailed human motion, sampling more than1000 unique points on the body . Our method outputs highly accurate 4D(spatio-temporal) point coordinates and, crucially, automatically assigns aunique label to each of the points . The locations and unique labels of thepoints are inferred from individual 2D input images only, without relying on any human body shape or skeletal kinematics models .…

## A Global to Local Double Embedding Method for Multi person Pose Estimation

Multi-person pose estimation is a fundamental and challenging problem to many computer vision tasks . Most existing methods can be broadly categorized intotwo classes: top-down and bottom-up methods . We present a novel method to simplify the pipeline by implementing person detection and joints detection simultaneously .…

## 3D Fully Convolutional Neural Networks with Intersection Over Union Loss for Crop Mapping from Multi Temporal Satellite Images

The proposed method was applied to identify soybean andcorn from a study area situated in the US corn belt using multi-temporal Landsat images . The study shows that our method outperforms related methods, . obtaining a Kappa coefficient of 90.8% .…

## On the Value of Wikipedia as a Gateway to the Web

By linking to external websites, Wikipedia can act as a gateway to the Web . In one month, English Wikipedia generated 43M clicks to external sites . Official links listed in infoboxes have by far the highestclick-through rate (CTR), 2.47% on average .…

## Vehicle to Vehicle V2V Communication Protocol Components Benefits Challenges Safety and Machine Learning Applications

Vehicle to vehicle communication is a new technology that enables vehicles onroads to communicate with each other to reduce traffic, accidents and ensure the safety of people . The main objective of vehicle-to-vehicle communicationprotocol is to create an effective communication system for intelligent transport systems .…

## A Tale of Three Datasets Towards Characterizing Mobile Broadband Access in the United States

Understanding and improving mobile broadband deployment is critical tobridging the digital divide and targeting future investments . In 2019, the Federal CommunicationsCommission (FCC) released a report on the progress of mobile broadbanddeployment in the United States . This report received a significant amount ofcriticism with claims that the cellular coverage, mainly available throughLong-Term Evolution (LTE), was over-reported in some areas, especially rural and/or tribal areas .…

## Decentralized Distributed Optimization for Saddle Point Problems

We consider distributed convex-concave saddle point problems over arbitrary connected undirected networks . We propose a decentralized distributed algorithm for their solution . The local functions distributed across the nodes are assumed to have global and local groups of variables .…

## Simulation based Optimization and Sensibility Analysis of MPI Applications Variability Matters

Finely tuning MPI applications and understanding the influence of keyparameters is critical . We demonstrate the capability of our method withHigh-PerformanceLinpack (HPL), the benchmark used to rank supercomputers intheTOP500, which requires careful tuning . We present an extensive (in)validation study that compares simulation with realexperiments and demonstrates our ability to predict the performance of HPL within a few percent consistently .…

## A first look into the carbon footprint of federated learning

Federated Learning (FL) is starting to be deployed at a global scale by companies that must adhere to new legal demands and policies for privacy protection . Despite impressive results, deep learning-based technologies also raisesevere privacy and environmental concerns . This paper offers the first-ever systematic study of the carbon footprint of FL .…

## Byzantine Dispersion on Graphs

The problem of Byzantine dispersion asks: given $n$ robots, up to $f$ of which are Byzantine, placed arbitrarily on an anonymous graph, design a terminating algorithm to be run by the robots . Each node has atmost one non-Byzantine robot on it .…

## GradPIM A Practical Processing in DRAM Architecture for Gradient Descent

GradPIM is an incremental, simple architectural design that does not invade the memory protocol . The proposed architecture can improve the performance of DNN training and greatly reduce memory bandwidth requirement while posing only a minimal amount of overhead to the protocol and DRAM area .…

## On the Impact of Device and Behavioral Heterogeneity in Federated Learning

Federated learning (FL) is becoming a popular paradigm for collaborativelearning over distributed, private datasets owned by non-trusting entities . FL comes with the challenge of performing training over largely heterogeneous datasets, devices, networks that are out of the control of the centralized FL server .…

## T RACKs A Faster Recovery Mechanism for TCP in Data Center Networks

TCP is oblivious to the composite nature of application data and artificially inflates the FCT of such flows by several orders of magnitude . This is due to TCP’s Internet-centric design that fixes theretransmission timeout (RTO) to be at least hundreds of milliseconds .…

## Large coverage fluctuations in Google Scholar a case study

This study analyses a large decrease incoverage of documents in the field of Astronomy and Astrophysics that took place in 2019 and its subsequent recovery, using longitudinal data from previous analyses and a new dataset extracted in 2020 . Disappeared documents did not reappear until the following index-wide update, many months after the problem was discovered .…

## Expansions in Cantor real bases

We introduce and study series expansions of real numbers with an arbitraryCantor real base . We pay special attention to periodic Cantor real bases, which we callalternate bases . We prove a generalization of Parry’s theorem characterizingsequences of nonnegative integers that are the greedy$\boldsymbol{\beta$-representations of some real number in the interval$[0,1)$.…

## Preimages under the Queuesort algorithm

Following the footprints of what have been done with the algorithm Stacksort, we investigate the preimages of the map associated with a slightly less wellknown algorithm, called Queuesort . We provide a recursion description of the set of all preimages, which can be also translated into a recursive procedure to find such preimages .…

## Dominance inequalities for scheduling around an unrestrictive common due date

The problem considered in this work consists in scheduling a set of tasks on a single machine, around an unrestrictive common due date . This problem can be formulated as a compact mixed integer program (MIP) In this article, we focus onneighborhood-based dominance properties, where the neighborhood is associatedto insert and swap operations .…

## The Phase Transition of Discrepancy in Random Hypergraphs

Motivated by the Beck-Fiala conjecture, we study the discrepancy problem intwo related models of random hypergraphs on $n$ vertices and $m$ edges . We show that withhigh probability (w.h.p.) $H_1$ has discrepancy at least $Omega(2-n/m) when $m = O(n)$ and $p=d/m$.…

## Local Access to Random Walks

For a graph $G$ on $n$ vertices, naively sampling the position of a randomwalk of at time $t$ requires work $Omega(t)$. We desire local accessalgorithms supporting $O}(G,s,t) queries, which return theposition of a . random walk from some start . $s$ at time of a walk, where the joint distribution of returned positions is $1/\text{poly}(n)$ close to theuniform distribution over such walks in $ell_1$ distance .…

## Dynamic Membership for Regular Languages

We study the dynamic membership problem for regular languages: fix a languageL, read a word w, build in time O(|w|) a data structure indicating if w is inL, and maintain this structure efficiently under substitution edits on w . We show that the problem is in O(log log |w| / log log | w|) for languages in analgebraically-defined class QSG .…

## Polynomial time algorithms in invariant theory for torus actions

An action of a group on a vector space partitions the latter into a set oforbits . We consider three natural and useful algorithmic “isomorphism” or “classification” problems . These capture and relate to a variety of problems within mathematics, physics and computer science, optimization andstatistics .…

## Fair and Optimal Cohort Selection for Linear Utilities

The rise of algorithmic decision-making has created an explosion of research around the fairness of those algorithms . The fair cohort selection problem captures a specific application where a single fair classifier is composed with itselfto pick a group of candidates of size exactly $k$.…

## Testing properties of signed graphs

In graph property testing the task is to distinguish whether a graphsatisfies a given property or is “far” from having that property . In this work we initiate the studyof property testing in signed graphs, where every edge has either a positive or negative sign .…

## Locality and Centrality The Variety ZG

We study the variety ZG of monoids where elements that belong to a group are central, i.e., commute with all other elements . We show that ZG is local,that is, the semidirect product ZG*D of ZG by definite semigroups is equal to LZG .…

## Learning Pomset Automata

We extend the L* algorithm to learn bimonoids recognising pomset languages . We then identify a class of automata that accepts precisely the class of pomsets languages recognised by pimonoids .…

## On synthesizing Skolem functions for first order logic formulae

Skolem functions play a central role in logic, from eliminating quantifiers in first order logic formulas to providing functional implementations ofrelational specifications . The question of how to effectively compute them is also important and useful for several applications . We show that even under reasonable assumptions on the signature of the formula, it is impossible to compute or synthesize SkoleM functions .…

## Model bounded monitoring of hybrid systems

Monitoring of hybrid systems attracts both scientific and practicalattention . We introduce a model-bounded monitoring scheme, where we useprior knowledge about the target system to prune interpolation candidates . We present two partial algorithms – one isvia reduction to reachability in LHAs and the other is a direct one usingpolyhedra – and show that these methods are efficient and practically relevant .…

## MatchKAT An Algebraic Foundation For Match Action

We present MatchKAT, an algebraic language for modeling match-action packetprocessing in network switches . We hope to embark on the first steps in exploring how networkprograms compiled to match-actions can be reasoned about formally in areliable, algebraic way . We alsodemonstrate the complexity of deciding equivalence in MatchKat isPSPACE-complete .…

## A NeRF Surface free Human 3D Pose Refinement via Neural Rendering

We propose a novel test-time optimizationapproach for monocular motion capture that learns a volumetric body model of the user in a self-supervised manner . Our proposed skeleton embedding serves as a common referencethat links constraints across time, thereby reducing the number of required camera views from traditionally dozens of calibrated cameras, down to a singleuncalibrated one .…

## An Overview of Agent based Traffic Simulators

Computer-based simulation is an accepted means for investigating the effects of new transportation policies and services . In most countries population in urban areas is growing, while availabletravel infrastructure and resources are limited . At the same time desires tominimise environmental impact and energy use have led to new requirements inthe field of inner-city transportation .…

## Tight Revenue Gaps among Multi Unit Mechanisms

This paper considers Bayesian revenue maximization in the $k$-unit setting . Four basicmechanisms among others have been widely employed in practice and widelystudied in the literature . We investigate the largest possible ratio between the two revenues (a.k.a.\ the revenue gap), overall possible value distributions of the buyers .…

## Selecting Matchings via Multiwinner Voting How Structure Defeats a Large Candidate Space

Given a set of agents with approval preferences over each other, we study the task of finding $k$ matchings fairly representing everyone’s preferences . We show that proportional approval voting (PAV), a well-establishedbut computationally intractable voting rule, becomes polynomial-timecomputable, and its sequential variant (seq-PAV) fulfills a rather strong guarantee known as extended justified representation .…

## Log time Prediction Markets for Interval Securities

We design a prediction market to recover a complete and fully general probability distribution over a random variable . Traders buy and sell intervalsecurities that pay $1 if the outcome falls into an interval and $0 otherwise . Our first design replicates the popular logarithmic market scoring rule (LMSR) but operates exponentially faster than a standard LMSR .…