## Exploring Simple Siamese Representation Learning

Siamese networks have become a common structure in various recent models for visual representation learning . These models maximize thesimilarity between two augmentations of one image, subject to certain conditions for avoiding collapsing solutions . Our experiments show thatcollapsing solutions do exist for the loss and structure, but a stop-gradientoperation plays an essential role in preventing collapsing .…

## Neural Scene Graphs for Dynamic Scenes

In this work, we present the first neural renderingmethod that decomposes dynamic scenes into scene graphs . We propose a learnedscene graph representation, which encodes object transformation and radiance, to efficiently render novel arrangements and views of the scene . We assessthe proposed method on synthetic and real automotive data, validating that ourapproach learns dynamic scenes – only by observing a video of this scene – andallows for rendering novel photo-realistic views of novel scene compositions with unseen sets of objects at unseen poses .…

## Filtering Rules for Flow Time Minimization in a Parallel Machine Scheduling Problem

This paper studies the scheduling of jobs of different families on parallel machines with qualification constraints . Originating from semiconductormanufacturing, this constraint imposes a time threshold between the execution of two jobs of the same family . Using this algorithm one can derivedfiltering rules on different variables of the model .…

## Collaborative Storytelling with Large scale Neural Language Models

Storytelling plays a central role in human socializing and entertainment . We present a collaborative storytelling system which works with a human storyteller to create a story by generating new utterances based on the story so far . We constructed the storytelling system by tuning apublicly-available large scale language model on a dataset of writing prompts and their accompanying fictional works .…

## Long Short Term Memory Networks for Bandwidth Forecasting in Mobile Broadband Networks under Mobility

Bandwidth forecasting in Mobile Broadband (MBB) networks is a challenging task, particularly when coupled with a degree of mobility . HINDSIGHT++ is an open-source R-based framework for bandwidthforecasting experimentation in MBB networks with Long Short Term Memory (LSTM)networks . The study aims to investigate the impact of hyperparameter optimization on achieving state-of-the-art performance and beyond .…

## MobileDepth Efficient Monocular Depth Prediction on Mobile Devices

Depth prediction is fundamental for many useful applications on computervision and robotic systems . On mobile phones, some useful applications such as augmented reality, autofocus and so on could be enhanced by accurate depth prediction . An Android application has been developed which can load CNN models topredict depth map by the monocular images captured from the mobile camera .…

## Sequential Defaulting in Financial Networks

We study financial networks, where banks are connected by contracts such as debts or credit default swaps . We analyze these networks in a sequential model where banks announce their default one at a time, and the system evolves in a step-by-step manner .…

## Modeling and Engineering Constrained Shortest Path Algorithms for Battery Electric Vehicles

We study the problem of computing constrained shortest paths for batteryelectric vehicles . Since battery capacities are limited, fastest routes are often infeasible . Users are interested in fast routes on which the energy consumption does not exceed the battery capacity .…

## Recovering the Imperfect Cell Segmentation in the Presence of Dynamically Localized Proteins

The method will empower microscopic experiments aimed at understanding molecular and cellular function . We demonstrate the value of this approach over frame-by-frame segmentation and regular temporal propagation on data from human embryonic kidney (HEK293T) cells transiently transfected with a fluorescent protein that moves in and out of the nucleus over time .…

## Survey and Open Problems in Privacy Preserving Knowledge Graph Merging Query Representation Completion and Applications

Knowledge Graph (KG) has attracted more and more companies’ attention for itsability to connect different types of data in meaningful ways . However, the data isolation problem limits the performance of KG and prevents its further development . Multiple parties have their own KGs but they cannot share with each other due to regulation or competition reasons .…

## Locally Aware Constrained Games on Networks

Network games have been instrumental in understanding strategic behavior of networks for applications such as critical infrastructure networks, social networks, and cyber-physical systems . One critical challenge of network games is that the behaviors of the players are constrained by the underlying physicallaws or safety rules, and the players may not have complete knowledge of network-wide constraints .…

## SophiaPop Experiments in Human AI Collaboration on Popular Music

A diverse team of engineers, artists, and algorithms, collaborated to createsongs for SophiaPop, via various neural networks, robotics technologies, andartistic tools . SophiaPop represents a human-AI collaboration, aspiring toward humanAI symbiosis . We believe that such a creative convergence of multipledisciplines with humans and AI working together, can make AI relevant to humanculture in new and exciting ways, and lead to a hopeful vision for the future of human- AI relations .…

## Exploring the political pulse of a country using data science tools

In this paper we consider tweets from leaders of political parties as a dynamical proxy to political programmes and ideas . We analyse levels of positive and negative sentiment in the tweetsusing new tools adapted to social media . We also train an Artificial Intelligence to recognise the political affiliation of a tweet .…

## Improvement of Classification in One Stage Detector

RetinaNet proposed Focal Loss for classification task and improved one-stagedetectors greatly . Most ofpredicted boxes whose IoU with ground-truth boxes are greater than 0.5, while their classification scores are lower than .5 . Our method can not only improve classification task, but also easemisalignment of classification and localization .…

## Meta Learning for Time Series Forecasting Ensemble

Meta-learning uses two separate random forest regression models, built on 390 timeseries features, to rank 22 univariate forecasting methods . Forecasting ensemble is formed from methods ranked as the best and forecasts are pooled using either simple or weighted average .…

## Bridging Scene Understanding and Task Execution with Flexible Simulation Environments

TESSE (Task Executionwith Semantic Segmentation Environments) is an open source simulator for developing scene understanding and task execution algorithms . It has been used to develop state-of-the-art solutions for metric-semantic mapping and 3Ddynamic scene graph generation . Code for the code is available at https://://://github.com/MIT-TESSE.…

## Hop Constrained Oblivious Routing

We prove the existence of an oblivious routing scheme that is $\mathrm{poly}(\log n)-competitive in terms of$congestion + dilation . The dilation is the maximum path hop-length, the congestion is the number of paths that include any single edge . The routing scheme obliviously and randomly selects a path for each packet independent of (the existence of) the other packets .…

## Image Denoising by Gaussian Patch Mixture Model and Low Rank Patches

A new method is proposed by solving two issues: how to improve similar patches matching accuracy and build anappropriate low rank matrix approximation model for Gaussian noise . Local patchmatching is to find similar patches in a large neighborhood which can alleviate effect, but the number of patches may be insufficient .…

## Fine Tuning BERT for Sentiment Analysis of Vietnamese Reviews

Sentiment analysis is an important task in the field ofNature LanguageProcessing (NLP) Many deep learning models have been proposed to tackle this task . BERT fine-tuning method produces amodel with better performancethan the original BERT model, according to the authors of this paper .…

## Cascade Attentive Dropout for Weakly Supervised Object Detection

Weakly supervised object detection (WSOD) aims to classify and locate objectswith only image-level supervision . We purposely discardattentive elements in both channel and space dimensions, and capture theinter-pixel and inter-channel dependencies to induce the model to better understand the global context .…

## SIMF Single Instruction Multiple Flush Mechanism for Processor Temporal Isolation

Asingle-instruction multiple-flush (SIMF) mechanism to flush the core-level state is used by operating-system-level solutions . SIMF significantly alleviates the overhead offlushing by more than a factor of two in execution time and reduces dynamic instruction count by orders-of-magnitude . The resultant processor isprototyped on Xilinx ZCU102 FPGA and validated with seL4microkernel, Linux kernel in multi-core scenarios, and a cache timing attack attack .…

## Self Supervised Small Soccer Player Detection and Tracking

State-of-the-art tracking algorithms achieve impressive results in scenarios on which they have beentrained for, but fail in challenging ones such as soccer games . Self-supervised pipeline is able to detect and track low-resolution soccer players under different recording conditions without any need of ground-truth data .…

## DeepPhaseCut Deep Relaxation in Phase for Unsupervised Fourier Phase Retrieval

Fourier phase retrieval is a classical problem of restoring a signal only from the measured magnitude of its Fourier transform . Modern methods such as PhaseLift and PhaseCut may offer performance guarantees with the help of convex relaxation . However, these algorithms are usuallycomputationally intensive for practical use .…

## Born Identity Network Multi way Counterfactual Map Generation to Explain a Classifier s Decision

There exists an apparent negative correlation between performance andinterpretability of deep learning models . We propose Born Identity Network (BIN) which is a post-hocapproach for producing multi-way counterfactual maps . Counterfactual maptransforms an input sample to be classified as a target label, similarto how humans process knowledge through counterfactsual thinking .…

## Deep Snapshot HDR Imaging Using Multi Exposure Color Filter Array

In this paper, we propose a deep snapshot high dynamic range (HDR) imaging framework that can effectively reconstruct an HDR image from the RAW datacaptured using a multi-exposure color filter array (ME-CFA) We introduce the idea of luminancenormalization that simultaneously enables effective loss computation and input data normalization by considering relative local contrasts in the”normalized-by-luminance” HDR domain .…

## Reconfigurable Intelligent Surface Enabled Federated Learning A Unified Communication Learning Design Approach

Federated learning (FL) has been proposed as an attractive substitute for centralized machine learning (ML) By collaboratively training a sharedlearning model at edge devices, FL avoids direct data transmission and thus avoids high communication latency and privacy issues as compared to centralized ML .…

## SLADE A Self Training Framework For Distance Metric Learning

Most existing distance metric learning approaches use fully labeled data to learn the sample similarities in an embedding space . We present a self-trainingframework, SLADE, to improve retrieval performance by leveraging additionalunlabeled data . We evaluate ourmethod on standard retrieval benchmarks: CUB-200, Cars-196 and In-shop.…

## Edge Adaptive Hybrid Regularization Model For Image Deblurring

A large parameter of regularization item reduces noise better in smooth area but blurs edges, while a small parameter sharpens edges but causes residual noise . Computationally, thenewly-established model is convex, then it can be solved by the semi-proximalalternating direction method of multipliers (sPADMM) with a linear-rateconvergence .…

## Virtues of Patience in Strategic Queuing Systems

A paper on the problem of selfish agents in discrete-time queuing systems shows that with no-regret learners, the system needs twice the capacity as would be required in the coordinated setting to ensure queue lengths remain stable despite the selfish behavior of the queues .…

## Graph Tikhonov Regularization and Interpolation via Random Spanning Forests

Novel Monte Carlo estimators are proposed to solve both the Tikhonovregularization (TR) and the interpolation problems on graphs . These estimatorsare based on random spanning forests (RSF), the theoretical properties of whichenable to analyze the estimators’ theoretical mean and variance .…

## Recovery to Efficiency A New Robustness Concept for Multi objective Optimization under Uncertainty

This paper presents a new robustness concept for uncertain multi-objectiveoptimization problems . Several approaches for generating recovery-to-efficiency robust sets are proposed as well . An extensive experimentalanalysis is performed to disclose differences among robust sets obtained using different concepts as well as to deduce some interesting observations .…

## Are Chess Discussions Racist An Adversarial Hate Speech Data Set

Antonio Radi\’c’s YouTube handle got blocked because it contained”harmful and dangerous” content . YouTube did not give further specific specific reason, but he speculated that given the current political situation, a referral to “black againstwhite” in the context of chess earned him this temporary ban .…

## DoDNet Learning to segment multi organ and tumors from multiple partially labeled datasets

A dynamicon-demand network (DoDNet) learns to segment multiple organs and tumors on partially labeled datasets . DoDNet consists of a shared encoder-decoderarchitecture, a task encoding module, a controller for generating dynamicconvolution filters, and a single but dynamic segmentation head . The information of the current segmentation task is encoded as a task-aware priorto tell the model what the model is expected to solve .…

## Unbalanced Incomplete Multi view Clustering via the Scheme of View Evolution Weak Views are Meat Strong Views do Eat

Incomplete multi-view clustering is an important technique to deal with real-world data . Different views often have distinct incompleteness, which results in strong views (low-incompleteness views) and weak views (high-incomplete views) The UnbalancedIncomplete Multi-view Clustering method (UIMC) improves the clustering performance by up to 40% on three evaluation metrics over other state-of-the-art methods .…

## 1st AfricaNLP Workshop Proceedings 2020

Proceedings of the 1st AfricaNLP Workshop held on 26th April alongside ICLR2020, Virtual Conference, Formerly Addis Ababa Ethiopia .…

## Planning Folding Motion with Simulation in the Loop Using Laser Forming Origami and Thermal Behaviors as an Example

Recent works have shown success in estimating foldability of adesign using robot motion planners . However, many foldable structures areactuated using physically coupled reactions (i.e., folding originated fromthermal, chemical, or electromagnetic loads) Therefore, a reliable foldabilityanalysis must consider additional constraints that resulted from these criticalphenomena .…

## Assessing out of domain generalization for robust building damage detection

An important step for limiting the negative impact of natural disasters israpid damage assessment after a disaster occurred . Building damage detection models have so far been evaluated mostly in thesimpler yet unrealistic in-distribution (IID) test setting . Here, we argue that future work should focus on the OOD regime instead .…

## On barren plateaus and cost function locality in variational quantum algorithms

A barren plateau is the phenomenon of exponentiallyvanishing gradients in sufficiently expressive quantum circuits . It has been established that the onset of a barren plateau regime depends on the cost function . Here we derive a lower bound on thevariance of the gradient, which depends mainly on the width of the circuitcausal cone of each term in the Pauli decomposition of a cost function.…

## On Error Exponents of Encoder Assisted Communication Systems

We consider a point-to-point communication system, where in addition to theencoder and the decoder, there is a helper that observes non-causally therealization of the noise vector and provides a (lossy) rate-$R$ description of it to the encoder . We study the cases of both fixed-rate and variable-rate noisedescriptions by the helper .…

## A Deep Language independent Network to analyze the impact of COVID 19 on the World via Sentiment Analysis

Wuhan experienced an outbreak of novel coronavirus, which soon spread all over the world, resulting in a deadly pandemic that infected millions of people around the globe . We propose a deep language-independent Multilevel Attention-basedConv-BiGRU network (MACBiG-Net), which includes embedding layer, word-levelencoded attention, and sentence-level encoded attention mechanism to extract positive, negative, and neutral sentiments .…

## ONION A Simple and Effective Defense Against Textual Backdoor Attacks

Backdoor attacks are a kind of emergent training-time threat to deepneural networks (DNNS) They can manipulate the output of DNNs and posses highinsidiousness . In this paper, we propose a simple and effective textualbackdoor defense named ONION, which is based on outlier word detection and might be the first method that can handle all the attack situations .…

## SalSum Saliency based Video Summarization using Generative Adversarial Networks

SalSum is based on a Generative Adversarial Network (GAN) model pre-trained with human eye fixations . The main contribution of the proposed method is that it provides perceptually compatible video summaries by combining both perceivedcolor and spatiotemporal visual attention cues in a unsupervised scheme .…

## Action Duration Prediction for Segment Level Alignment of Weakly Labeled Videos

This paper focuses on weakly-supervised action alignment, where only the sequence of video-level actions is available for training . We propose anovel Duration Network, which captures a short temporal window of the video andlearns to predict the remaining duration of a given action at any point in timewith a level of granularity based on the type of that action .…

## A lightweight cryptography LWC framework to secure memory heap in Internet of Things

Java is the most common platform for embedded applications such as IoT, Wireless Sensors Networks (WSN), Near Field Communications (NFC) and RFID . Object programming languagessuch as Java, SWIFT, PHP and C++ use garbage collection after any object run which creates security loophole for attacks such as Next Memory AddressOccupation (NMAO), memory replay, Learning Tasks Behaviors (LTB) The proposed method prevents directed attack by encrypting the objectGarbage Collection at run time .…

## PSD2 Explainable AI Model for Credit Scoring

The aim of this paper is to develop and test advanced analytical methods to improve the prediction accuracy of Credit Risk Models, preserving at the sametime the model interpretability . The project focuses on applying an explainable machine learning model to PSD2-related databases .…

## PIFE PIC Parallel Immersed Finite Element Particle In Cell For 3 D Kinetic Simulations of Plasma Material Interactions

PIFE-PIC is a novel three-dimensional (3-D) ParallelImmersed-Finite-Element (IFE) Particle-in-Cell (PIC) simulation model . A simulation of theorbital-motion-limited (OML) sheath of a dielectric sphere immersed in astationary plasma is carried out to validate the code package . A large-scale simulation ofplasma charging at a lunar crater containing 2 million PIC cells (10 millionFE/IFE cells) and about 520 million particles, running for 20,000 PIC steps in about 109 wall-clock hours, is presented to demonstrate the high-performance computing capability of Pife-Pic .…

## Topic modelling discourse dynamics in historical newspapers

This paper addresses methodological issues in diachronic data analysis for historical research . We apply two families of topic models (LDA and DTM) on arelatively large set of historical newspapers . Our case study focuses on newspapers and periodicals published in Finland between 1854 and 1917 .…

## Weighted automata are compact and actively learnable

We show that weighted automata over the field of two elements can beexponentially more compact than non-deterministic finite state automata . We include an algorithm for learning WAs over any fieldbased on a linear algebraic generalization of the Angluin-Schapire algorithm .…

## User and Item aware Estimation of Review Helpfulness

In online review sites, the analysis of user feedback for assessing itshelpfulness for decision-making is usually carried out by studying the properties of individual reviews . We propose a novel helpfulness estimation model that extends previous ones with analysis of deviations in rating, length and polarity with respect to the reviews written by the same person, or concerning the same item .…

## A modified bond model for describing isotropic linear elastic material behaviour with the particle method

A modified bond model that includes the coupling of shear strain energy of neighbouring bonds is proposed . The coupling is described by a multi-bond term that enables the model to distinguish between shear deformations and rigid-body rotations . The positive definiteness of the strain energy function of the modified Bond model is verified.…