## Retrieval Augmentation to Improve Robustness and Interpretability of Deep Neural Networks

Deep neural network models have achieved state-of-the-art results in varioustasks related to vision and/or language . Most models are trained by iterating over single input-output pairs . We exploit the training data to improve the robustness andinterpretability of deep neural networks .…

## libtxsize a library for automated Bitcoin transaction size estimates

This paper presents libtxsize, a library to estimate the size requirements ofarbitrary Bitcoin transactions . To account for different use cases, the library provides estimates in bytes, virtual bytes, and weight units . The paper can also serve as reference for different Bitcoin data and transaction types, their semantics, and their size requirements .…

## LET Linguistic Knowledge Enhanced Graph Transformer for Chinese Short Text Matching

Chinese short text matching is a fundamental task in natural language processing . Existing approaches usually take Chinese characters or words as tokens . We introduce HowNet as an externalknowledge base and propose a Linguistic knowledge Enhanced graph Transformer . We adopt the word latticegraph as input to maintain multi-granularity information .…

## Emotion Aware Emotion Agnostic or Automatic Corpus Creation Strategies to Obtain Cognitive Event Appraisal Annotations

Appraisal theories explain how the cognitive evaluation of an event leads to a particular emotion . Smith andEllsworth (1985) showed that the appraisal dimensions attention, certainty,anticipated effort, pleasantness, responsibility/control and situationalcontrol discriminate between (at least) 15 emotion classes . We analyze two manual annotation settings: (1)showing the text to annotate while masking the experienced emotion label; (2) revealing the emotion associated with the text .…

## FAITH Fast iterative half plane focus of expansion estimation using event based optic flow

Course estimation is a key component for the development of autonomousnavigation systems for robots . While state-of-the-art methods widely use visual-based algorithms, it is worth noting that they all fail to deal with the complexity of the real world by being computationally greedy and sometimes tooslow .…

## Optimized Memoryless Fair Share HPC Resources Scheduling using Transparent Checkpoint Restart Preemption

Common resource management methods in supercomputing systems usually include hard divisions, capping, and quota allotment . Those methods involve bad supply-and-demand management rather than a free market playground that will eventually increase system utilization and productivity . In this work, we propose the newly Optimized Memoryless Fair-Share HPCResources Scheduling using Transparent Checkpoint-Restart Preemption, in which the social welfare increases using a free-of-cost interchangeable proprietarypossession scheme .…

## Blockchained Federated Learning for Threat Defense

Security systems fail to detect serious threats such as zero-day attacks . The need for more active and more effective security methods keeps increasing . The proposed framework combinesFederated Learning for the distributed and continuously validated learning of the tracing algorithms .…

## Approximate Privacy Preserving Neighbourhood Estimations

Anonymous social networks present a number of new and challenging problems for existing Social Network Analysis techniques . Traditionally, existing methods for analysing graph structure, such as community detection, required global knowledge of the graph structure . Exchanging this data structure infuture decentralised learning deployments gives away no information about theneighbours of the node and therefore does preserve the privacy .…

## Emerging Trends in Federated Learning From Model Fusion to Federated X Learning

Federated learning is a new learning paradigm that decouples data collection and model training via multi-party computation and model aggregation . As aflexible learning setting, federated learning has the potential to integrate with other learning frameworks . This survey reviews the state of the art, challenges, and future directions of learning algorithms .…

## Do Input Gradients Highlight Discriminative Features

In this work, we introduce an evaluation framework to study this hypothesis for benchmark image classification tasks . We make two surprising observations on CIFAR-10 and Imagenet-10 datasets . We introduce a synthetic testbed and theoretically justify our counter-intuitive empirical findings .…

## Toward Instance Optimal State Certification With Incoherent Measurements

We revisit the basic problem of quantum state certification . Given copies of unknown mixed state and description of amixed state $rho$ and $Omega(d^{\Theta(1)/\epsilon^2)$ copies are necessary . The exact exponent depends on the type of measurements the learner can make [OW15, BCL20] We give the first bounds of this nature for the quantum setting .…

## Benchmarking and Survey of Explanation Methods for Black Box Models

The widespread adoption of black-box models in Artificial Intelligence has increased the need for explanation methods to reveal how these obscure models reach specific decisions . We provide acategorization of explanation methods based on the type of explanation returned . We present the most recent and widely used explainers, and we show avisual comparison among explanations and a quantitative benchmarking .…

## Dual MINE based Neural Secure Communications under Gaussian Wiretap Channel

Some researches are devoted to the topic of end-to-end learning a physical layer secure communication system based on autoencoder under Gaussianwiretap channel . However, in those works, the reliability and security of theencoder model were learned through necessary decoding outputs of not onlylegitimate receiver but also the eavesdropper .…

## Distributionally Robust Federated Averaging

In this paper, we study communication efficient distributed algorithms fordistributionally robust federated learning via periodic averaging with adaptivesampling . We propose a Distributionally Robust Federated Federated Averaging(DRFA) algorithm that employs a novel snapshotting scheme to approximate theaccumulation of history gradients of the mixing parameter .…

## Performance Comparison for Scientific Computations on the Edge via Relative Performance

In a typical Internet-of-Things setting that involves scientificapplications, a target computation can be evaluated in many different ways . In this paper, we focus on analyzing theperformance of a given set of algorithms by clustering them into performanceclasses . We use a measurement-based approach to evaluate and scorealgorithms based on pair-wise comparisons .…

## Reducing Labelled Data Requirement for Pneumonia Segmentation using Image Augmentations

Deep learning semantic segmentation algorithms can localise abnormalities oropacities from chest radiographs . However the task of collecting andannotating training data is expensive and requires expertise which remains abottleneck for algorithm performance . We investigate the effect of imageaugmentations on reducing the requirement of labelled data in the semanticsegmentation of chest X-rays for pneumonia detection .…

## On Instabilities of Conventional Multi Coil MRI Reconstruction to Small Adverserial Perturbations

Deep learning (DL) has received much attention in accelerated MRI, but recent studies suggest small perturbations may lead to instabilities inDL-based reconstructions . However, these works focus on single-coil acquisitions, which is not practical . We investigate instabilities caused by small adversarial attacks for multi-coilsacquisitions .…

## Phragmén s Voting Methods and Justified Representation

Swedish mathematician Lars Edvard Phragm\'{e}n proposed a load-balancing approach for selecting committees based on approvalballots . We show that the sequential variant satisfiesproportional justified representation, which is a rare property for committeemonotonic methods . We also analyze the computational complexity of the methods and provide mixed-integer programming based algorithms for computing them.…

## Generalized Parametric Path Problems

Parametric path problems arise independently in diverse domains, ranging from transportation to finance, where they are studied under various assumptions . We show that when the parametric weights are linear, algorithms remain tractable even under relaxed assumptions . If the weights are allowed to be non-linear, the problem becomes NP-hard.…

## Iterative Bounding MDPs Learning Interpretable Policies via Non Interpretable Methods

Current work in explainable reinforcement learning generally produces policies in the form of a decision tree over the state space . We propose a novel Markov Decision Process (MDP) type for learning decision tree policies: Iterative Bounding MDPs (IBMDPs) An IBMDP is constructed around a base MDP so each IBMDP policy is guaranteed to correspond to a .…

## ZJUKLAB at SemEval 2021 Task 4 Negative Augmentation with Language Model for Reading Comprehension of Abstract Meaning

This paper presents our systems for the three Subtasks of SemEval Task4:Reading Comprehension of Abstract Meaning (ReCAM) We explain the algorithms used to learn our models and the process of tuning the algorithms and selecting the best model . Our models achieve the 4th rank on both official test sets of Subtask 1 and Subtask 2 with an accuracy of 87.9% and 92.8% .…

## QNLP in Practice Running Compositional Models of Meaning on a Quantum Computer

Quantum Natural Language Processing (QNLP) deals with the design and implementation of NLP models intended to be run on quantum hardware . In thispaper, we present results on the first NLP experiments conducted on NoisyIntermediate-Scale Quantum (NISQ) computers for datasets of size < 100sentences . Exploiting the formal similarity of the compositional model of meaning by Coecke et al. with quantum theory, we create representationsfor sentences that have a natural mapping to quantum circuits . We use theserepresentations to implement and successfully train two models that solvesimple sentence classification tasks . …

## Subcubic Certificates for CFL Reachability

Many problems in interprocedural program analysis can be modeled as the context-free language (CFL) reachability problem . Despite years of efforts, there are no known truly sub-cubical algorithms for this problem . We study the related certification task: given aninstance of CFL reachability, are there small and efficiently checkablecertificates for the existence and for the non-existence of a path?…

Smart speakers and voice-based virtual assistants are vulnerable to various privacy threats exploiting machine learning to analyze the generated encrypted traffic . To cope with that, deep adversarial learning approaches can be used tobuild black-box countermeasures altering the network traffic (e.g.,…

## Imitation Learning for Robust and Safe Real time Motion Planning A Contraction Theory Approach

Learning-based Autonomous Guidance with Robustness,Optimality, and Safety guarantees (LAG-ROS) is a real-time robust motion planning algorithm for safety-critical nonlinear systems perturbed by boundeddisturbances . For the CLF, we exploit a neural-network-based method of NeuralContraction Metrics (NCMs), which provides a differential Lyapunov function to minimize an upper bound of the steady-state Euclidean distance betweenperturbed and unperturbed system trajectories .…

## Coalgebra Encoding for Efficient Minimization

Recently, we have developed an efficient generic partition refinementalgorithm, which computes behavioural equivalence on a state-based system givenas an encoded coalgebra, and implemented it in the tool CoPaR . Here we extendthis to a fully fledged minimization algorithm and tool by integrating two new aspects: (1) the computation of the transition structure on the minimized stateset, and (2) the computations of the reachable part of the given system .…

## Non invasive Cognitive level Human Interfacing for the Robotic Restoration of Reaching Grasping

The system is tested with 5 healthy participants, showing an average successrate of $96.6\%$ on first attempt across 6 tasks . The interface is worn, calibrated and ready to use within 5 minutes . Users learn to control and make successful use of the system with an additional 5 minutes of interaction .…

## Scene Retrieval for Contextual Visual Mapping

Visual navigation localizes a query place image against a reference database of place images . Four different scene classes, including pedestrian crossings and stations, are identified in each of the Nordland andSt. Lucia datasets . Scene retrieval extends imageretrieval to classification of scenes defined at test time by associating asingle query image to reference images of scene classes .…

## Combinatorial Bandits under Strategic Manipulations

We study combinatorial multi-armed bandits (CMAB) understrategic manipulations of rewards . We propose a strategic variant of thecombinatorial UCB algorithm, which has a regret of at most $O(m\log T + mB_{max)$ under strategic manipulations . We further providelower bounds on the strategic budgets for attackers to incur certain regret of the bandit algorithm .…

## Swivel Hardening WebAssembly against Spectre

Swivel is a new compiler framework for hardening WebAssembly (Wasm) against Spectre attacks . Wasm has become a popular lightweight, in-process sandbox and is used in production to isolate different clients on edge clouds and function-as-a-service platforms . Spectre attacks can bypass Wasm’s isolation guarantees .…

## Algorithms and Complexity on Indexing Founder Graphs

We introduce a compact pangenome representation based on an optimalsegmentation concept that aims to reconstruct founder sequences from a multiplesequence alignment (MSA) Such founder sequences have the feature that each row of the MSA is a recombination of the founders .…

## An Advection Diffusion based Filter for Machinable Designs in Topology Optimization

This paper introduces a simple formulation for topology optimization problems . The method distinguishes itself from existing methods by using the advection-diffusion equation with Robin boundary conditions to perform a filtering of the design variables . The proposedapproach is less computationally expensive than the traditional methods used .…

## Maximizing Cosine Similarity Between Spatial Features for Unsupervised Domain Adaptation in Semantic Segmentation

A segmentation network mainly consists of two parts, a feature extractor and a classificationhead . We propose a novel method that tackles the problem of unsupervised domainadaptation for semantic segmentation . Our method computes a cosine similarity matrix between thesource feature map and the target feature map .…

## File fragment recognition based on content and statistical features

The known files are divided into different fragments, and different classificational algorithms are used to solve the problems of file fragment recognition . The proposed recognition algorithm can recognize 6 types of useful files and may distinguish a type of file fragments with higher accuracy thanthe similar works done .…

## Discrete Distribution Estimation with Local Differential Privacy A Comparative Analysis

Local differential privacy is a promising privacy-preserving model for statistical aggregation of user data that prevents user privacy leakage from the data aggregator . The Basic RAPPOR algorithm generally performs best for the benchmark datasets in the high privacy regime .…

## Spanish Biomedical and Clinical Language Embeddings

We computed both Word and Sub-word Embeddings using FastText . For Sub-words we selected Byte Pair Encoding (BPE) algorithm to represent thesub-words . We evaluated the Biomedical Word Embedding obtaining better resultsthan previous versions .…

## Graph Exploration by Energy Sharing Mobile Agents

We consider the problem of collective exploration of a known $n$-nodeedge-weighted graph by mobile agents that have limited energy but arecapable of energy transfers . The goal of the exploration problem is for everyedge in the graph to be traversed by at least one agent .…

## Sentiment Analysis of Persian English Code mixed Texts

The rapid production of data on the internet and the need to understand howusers are feeling from a business and research perspective has prompted the need for an automatic monolingual sentiment detection systems . We then introduce a model which uses BERT pretrained embeddings as well as translation models to automaticallylearn the polarity scores of these Tweets .…

## A Survey of RDF Stores SPARQL Engines for Querying Knowledge Graphs

Recent years have seen the growing adoption of non-relational data models forrepresenting diverse, incomplete data . The RDF graph-based datamodel has seen ever-broadening adoption, particularly on the Web . This adoption prompted the standardization of the SPARQL query language for RDF, as well as the development of a variety of local and distributed engines for processing queries over RDF graphs .…

## A New Neuromorphic Computing Approach for Epileptic Seizure Prediction

Several high specificity and sensitivity seizure prediction methods withconvolutional neural networks (CNNs) are reported . CNNs arecomputationally expensive and power hungry. These inconveniences make CNN-basedmethods hard to be implemented on wearable devices . A neuromorphic computing approach for seizure prediction is proposed in this work .…

## A Thin Self Stabilizing Asynchronous Unison Algorithm with Applications to Fault Tolerant Biological Networks

The Stone Age (SA) model provides an abstraction for network algorithms distributed overrandomized finite state machines . This model is designed to resemble the dynamic processes in cellular networks . The SA model assumes a weak communicationscheme that is built upon the nodes ability to sense their vicinity in anasynchronous manner .…

## Machine Learning Based Optimal Mesh Generation in Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) is a major sub-field of engineering . Finding an optimal mesh is key for computational efficiency . The proposed concept is validated along 2d wind tunnel simulations with more than 60,000 simulations . Corresponding predictions of optimal meshes can be used as input for any meshgeneration and CFD tool.…

## A Framework for Integrating Gesture Generation Models into Interactive Conversational Agents

Embodied conversational agents (ECAs) benefit from non-verbal behavior for efficient interaction with users . Gesticulation – hand and armmovements accompanying speech – is an essential part of non-verbial behavior . To date, recent end-to-end gesture generation methods have not been evaluated in a real-time interaction .…

## Are pre trained text representations useful for multilingual and multi dimensional language proficiency modeling

Development of language proficiency models for non-native learners has been an active area of interest in NLP research for the past few years . We report experiments with three languages — German, Italian, and Czech — and model seven dimensions of proficiency .…

## Blocks World Revisited The Effect of Self Occlusion on Classification by Convolutional Neural Networks

Despite recent successes in computer vision, there remain new avenues to explore . With TEOS (The Effect ofSelf-Occlusion), we propose a 3D blocks world dataset that focuses on thegeometric shape of 3D objects and their omnipresent challenge of self-occlusion . We present a dataset of twodifficulty levels (L1 and L2 ), containing 36 and 12 objects, respectively .…