Unsupervised Sentence embeddings by Manifold Approximation and Projection

The concept of unsupervised universal sentence encoders has gained traction recently, wherein pre-trained models generate effective task-agnosticfixed-dimensional representations for phrases, sentences and paragraphs . In thiswork we propose a novel technique to generate sentence-embeddings in anunsupervised fashion by projecting the sentences onto a fixed-dimensionalmanifold with the objective of preserving local neighbourhoods in the originalspace .…

CSS LM A Contrastive Framework for Semi supervised Fine tuning of Pre trained Language Models

Fine-tuning pre-trained language models (PLMs) has demonstrated itseffectiveness on various downstream NLP tasks . We introduce a novel framework (named “CSS-LM”) to improve the fine-tuneing phase of PLMs via contrastive semi-supervised learning . CSS-LM achieves better results than the conventional fine-tuned strategy on a series of downstream tasks with few-shotsettings, and outperforms the latest supervised contrastive language-model strategies .…

Blockchain Gateways Bridges and Delegated Hash Locks

In the current work we discuss the notion of gateways as a means forinteroperability across different blockchain systems . We discuss two keyprinciples for the design of gateway nodes and scalable gateway protocols . We illustrate the need for a standard gateway protocol by describing a unidirectional asset movement protocol between two peer gateways, under the strict condition of both blockchains being private/permissioned with their ledgers inaccessible to external entities .…

Privacy preserving Cloud based DNN Inference

Deep learning as a service (DLaaS) has been intensively studied to facilitatethe wider deployment of the emerging deep learning applications . However, DLaaSmay compromise the privacy of both clients and cloud servers . In this paper, we propose anovel privacy preserving cloud-based DNN inference framework (namely, “PROUD”), which greatly improves the computational efficiency .…

Robust Explanations for Private Support Vector Machines

We consider counterfactual explanations for private support vector machines, where the privacy mechanism that publicly releases the classifierguarantees differential privacy . While privacy preservation is essential whendealing with sensitive data, there is a consequent degradation in the classification accuracy due to the introduced perturbations in classifierweights .…

Fairness in ERC token markets A Case Study of CryptoKitties

A gene determination algorithm in CryptoKitties has little randomness, and a significant advantage to gain profit is givento players who know its bias over those who do not . We state incompleteness andimpact of the algorithm and other factors. Besides, we suppose countermeasuresto reduce CryptoKITTies’ unfairness as a market.…

Iconographic Image Captioning for Artworks

Image captioning implies automatically generating textual descriptions of images based only on the visual input . Not many contributions have been made in the domain of art historical data . This work aims to address some of those challenges by utilizing a novel large-scaledataset of artwork images annotated with concepts from the Iconclassclassification system designed for art and iconography .…

AttributeNet Attribute Enhanced Vehicle Re Identification

Vehicle Re-Identification (V-ReID) is a critical task that associates thesame vehicle across images from different camera viewpoints . We propose a new method to efficiently explorediscriminative information from vehicle attributes . We enable the interaction by distilling theReID-helpful attribute feature and adding it into the general ReID feature to increase the discrimination power.…

Self supervised driven consistency training for annotation efficient histopathology image analysis

Training a neural network with a large labeled dataset is still a dominantigm in computational histopathology . However, obtaining such exhaustivemanual annotations is often expensive, laborious, and prone to inter and intra-observer variability . We overcome this challenge by leveraging both task-agnostic and task-specific unlabeled databased on novel strategies: i) a self-supervised pretext task that harnesses the underlying multi-resolution contextual cues in histology whole-slide imagesto learn a powerful supervisory signal for unsupervised representationlearning .…

Damage detection using in domain and cross domain transfer learning

We investigate the capabilities of transfer learning in the area of structural health monitoring . Typical image datasets for such problems arerelatively small, calling for the transfer of learned representation from arelated large-scale datasets . We propose a combination of in-domain andcross-domain transfer learning strategies for damage detection in bridges .…

A novel multiple instance learning framework for COVID 19 severity assessment via data augmentation and self supervised learning

How to fast and accurately assess the severity level of COVID-19 is anessential problem, when millions are suffering from the pandemic around the world . Currently, the chest CT is regarded as a popular andinformative imaging tool for diagnosis . Weak annotation and insufficient data that may obstruct automatic COID-19 severity assessment with CT images .…

Multivariate Analysis of Scheduling Fair Competitions

A fair competition is anon-eliminating competition where each contestant (team or individual player)may not play against all other contestants . The total difficulty for each contestant is the same: the sum of the initial rankings of the opponents foreach . The winner of thefair competition is the contestant who wins the most games .…

Supporting Serendipity Opportunities and Challenges for Human AI Collaboration in Qualitative Analysis

Qualitative inductive methods are widely used in CSCW and HCI research for their ability to generatively discover deep and contextualized insights . But these inherently manual and human-resource-intensive processes are oftenfeasible for analyzing large corpora . Researchers have been increasingly interested in ways to apply qualitative methods to “big” data problems, hoping to achieve more generalizable results from larger amounts of data while preserving the depth and richness of qualitative methods .…

Linking Labs Interconnecting Experimental Environments

LabLinking is a technology-based interconnection of experimental laboratories across institutions, disciplines, cultures, cultures and languages . In other words experiments without borders . Linked labs provide a platform for a continuous exchange between scientists and experimenters, argues the authors . Linking labs enable a time synchronous execution of experiments performed with and by decentralized user and researchers, improving outreach and ease of subjectrecruitment, allowing to establish new experimental designs and to incorporate a panoply of complementary biosensors, devices, hard- and software solutions .…

Improving Accuracy and Diversity in Matching of Recommendation with Diversified Preference Network

Heterogeneous graph neuralnetwork framework for diversified recommendation (GraphDR) in matching . GraphDR builds a huge heterogeneous preference network to record different types of preferences, and conduct a field-level heterogeneous graph attentionnetwork for node aggregation. GraphDR has been deployed on a well-knownrecommendation system, which affects millions of users.…

Role of Attentive History Selection in Conversational Information Seeking

The rise of intelligent assistant systems like Siri and Alexa have led to the emergence of Conversational Search . We propose to use another history selection approach thatynamically selects and weighs history turns using the attention mechanism forquestion embedding . The novelty of our approach lies in experimenting with softattention-based history selection approaches in an open-retrieval setting.…

Coded Computing with Noise

Distributed computation is a framework used to break down a complex task into smaller tasks and distributing them among computational nodes . Erasure correction codes have recently been introduced and have become apopular workaround to the well known “straggling nodes” problem .…

An Autonomous Negotiating Agent Framework with Reinforcement Learning Based Strategies and Adaptive Strategy Switching Mechanism

Autonomous Negotiating Agent Framework allows real-time classification of opponent’s behaviour . Framework has reviewer component which enables self-enhancement capability by deciding to include new strategies or replace old ones with better strategies periodically . We demonstrate an instance of our framework by implementing maximum entropy reinforcement learning basedstrategies with a deep learning based opponent classifier .…

Mesh robustness of the variable steps BDF2 method for the Cahn Hilliard model

The two-step backward differential formula (BDF2) implicit method withunequal time-steps is investigated for the Cahn-Hilliard model . The suggested method isproved to preserve a modified energy dissipation law at the discrete levels . We view the BDF2 formula as aconvolution approximation of the first time derivative and perform the erroranalysis by using the recent suggested discrete orthogonal convolution kernels .…

Dynamic Movement Primitives in Robotics A Tutorial Survey

Dynamic Movement Primitives (DMPs) represent an elegant mathematicalformulation of the motor primitives as stable dynamical systems . They are wellsuited to generate motor commands for artificial systems like robots . DMPs have inspired researchers in different robotic fields including imitation and reinforcement learning, optimal control, and human-robot co-working, resulting a considerable amount of published papers .…

Embedding manifold structures into Kalman filters

Error-state Kalman filter is an elegant and effective filtering technique forrobotic systems operating on manifolds . To avoid tedious and repetitivederivations for implementing an error-state filter for a certain system, this paper proposes a generic symbolic representation . This is particularly useful when the robotic system is of highdimension .…

Lightweight 3 D Localization and Mapping for Solid State LiDAR

LiDAR sensor has become one of the most important perceptual devices due to its important role in simultaneouslocalization and mapping (SLAM) Existing SLAM methods are mainly developed formechanical LiDar sensors, which are often adopted by large scale robots . We propose a new SLAM framework for solid-stateLiDAR sensors which involves feature extraction, odometryestimation, and probability map building .…