Prior Signal Editing for Graph Filter Posterior Fairness Constraints

Graph filters are an emerging paradigm that systematizes informationpropagation in graphs as transformation of prior node values, called graphsignals, to posterior scores . In this work, we study the problem of mitigatingdisparate impact, i.e. posterior score differences between a protected set ofsensitive nodes and the rest, while minimally editing scores to preserverecommendation quality .…

Quantum Sub Gaussian Mean Estimator

We present a new quantum algorithm for estimating the mean of a real-valued random variable obtained as the output of a quantum computation . Our estimatorachieves a nearly-optimal quadratic speedup over the number of classical i.i.d.samples needed . We obtain new quantum algorithms for the .…

Representation and Processing of Instantaneous and Durative Temporal Phenomena

In this paper, we propose a new logic based temporal phenomenadefinition language specifically tailored for Complex Event Processing . We demonstrate the expressiveness ofour proposed language by employing a maritime use case where we define maritimeevents of interest . Finally, we analyse the execution semantics of our proposed language for stream processing and introduce the `Phenesthe’ implementationprototype.…

Quantum Sub Gaussian Mean Estimator

We present a new quantum algorithm for estimating the mean of a real-valued random variable obtained as the output of a quantum computation . Our estimatorachieves a nearly-optimal quadratic speedup over the number of classical i.i.d.samples needed . We obtain new quantum algorithms for the .…

Learning to Give Checkable Answers with Prover Verifier Games

Prover-Verifier Games (PVGs) is a game-theoretic framework to encourage learningagents to solve decision problems in a verifiable manner . The PVG consists of two learners with competing objectives: a trusted verifier network tries tochoose the correct answer, and a more powerful but untrusted prover network attempts to persuade the verifier of a particular answer .…

4 bit Quantization of LSTM based Speech Recognition Models

We investigate impact of aggressive low-precision representations ofweights and activations in large LSTM-based architectures forAutomatic Speech Recognition (ASR) Using a 4-bit integer representation, a quantization approach results in significant Word ErrorRate (WER) degradation . We show that minimal accuracy loss is achievable with an appropriate choice of quantizers and initializations .…

Injecting Text in Self Supervised Speech Pretraining

The proposed method, tts4pretrain, complements the power of contrastivelearning in self-supervision with linguistic/lexical representations derived from synthesized speech . Lexical learning in the speech encoder is enforced through anadditional sequence loss term that is coupled with contrastive loss duringpretraining . We demonstrate that this novel pretraining method yields WordError Rate (WER) reductions of 10% relative on the well-benchmarked,Librispeech task over a state-of-the-art baseline pretrained with wav2vec2.0only .…

COVID 19 reproduction number estimated from SEIR model association with people s mobility in 2020

This paper is an exploratory study of two epidemiological questions on aworldwide basis . How fast is the disease spreading? Are the restrictions(especially mobility restrictions) for people bring the expected effect? To answer the first question, we propose a tool for estimating the reproductionnumber of epidemic (the number of secondary infections $R_t$) based on the SEIR model .…

Robustness Disparities in Commercial Face Detection

Facial detection and analysis systems have been deployed by large companies and critiqued by scholars and activists for the past decade . We present the first of its kind detailedbenchmark of the robustness of three such systems: Amazon Rekognition,Microsoft Azure, and Google Cloud Platform .…

SPARROW A Novel Covert Communication Scheme Exploiting Broadcast Signals in LTE 5G Beyond

This work proposes a novel framework to identify and exploit vulnerable MAClayer procedures in commercial wireless technologies for covert communication . Examples of covert communication include data exfiltration, remotecommand-and-control (CnC) and espionage . In this framework, the SPARROW schemesuse the broadcast power of incumbent wireless networks to covertly relay messages across a long distance without connecting to them .…

Canoe A System for Collaborative Learning for Neural Nets

Canoe is a framework that facilitates knowledgetransfer for neural networks . Canoe provides new system support for dynamicallyextracting significant parameters from a helper node’s neural network . The evaluation of Canoe with different PyTorchand TensorFlow neural network models demonstrates that the knowledge transfermechanism improves the model’s adaptiveness to changes up to 3.5X compared tolearning in isolation, while affording several magnitudes reduction in datamovement costs compared to federated learning compared to learning in isolation .…

A High Fidelity Flow Solver for Unstructured Meshes on Field Programmable Gate Arrays

In this work, we design a custom FPGA-based accelerator for a computationalfluid dynamics (CFD) code . We target the entire unstructured Poisson solver . We propose a novel datamovement-reducing technique where we compute geometric factors on the fly, which yields significant (700+ GFlop/s) single-precision performance and anupwards of 2x reduction in runtime for the local evaluation of the Laplace operator .…

Enel Context Aware Dynamic Scaling of Distributed Dataflow Jobs using Graph Propagation

Enel is a novel dynamic scaling approach that uses messagepropagation on an attributed graph to model dataflow jobs and, thus, allows forderiving effective rescaling decisions . Enel incorporates descriptiveproperties that capture the respective execution context, considers statisticsfrom individual dataflow tasks, and propagates predictions through the jobgraph to eventually find an optimized new scale-out .…

Cross domain Single channel Speech Enhancement Model with Bi projection Fusion Module for Noise robust ASR

In recent decades, studies have suggested that phase information is crucial for speech enhancement (SE) We put forward a novel cross-domainspeech enhancement model and a bi-projection fusion (BPF) mechanism fornoise-robust ASR . To evaluate the effectiveness of our proposed method, we conduct an extensive set of experiments on the publicly-available Aishell-1Mandarin benchmark speech corpus .…

Web Image Context Extraction with Graph Neural Networks and Sentence Embeddings on the DOM tree

Web Image Context Extraction (WICE) consists in obtaining the textual information describing an image using the content of the surrounding webpage . When done at a large scale (e.g., for search engine indexation), it may become very computationally costly (up to several seconds per page) To avoid this cost, we introduce a novel WICE approach that combines Graph NeuralNetworks (GNNs) and Natural Language Processing models .…

Optimal area visibility representations of outer 1 plane graphs

This paper studies optimal-area visibility representations of $n$-vertexouter-1-plane graphs . It claims that any graph of this family admits an embedding-preserving visibilityrepresentation whose area is $O(n^{1.5) The paper claims that this area bound is worst-case optimal . We also extend the study to other representation models and, among other results, constructasymptotically optimal $o(n\, pw(G),$ is the pathwidth of the outer-1 planar graph $G$ The paper concludes that the optimal area bound can be achieved if we do not respect the embedding but still have at most one crossing per edge .…

Understanding Money Trails of Suspicious Activities in a cryptocurrency based Blockchain

The decentralization, redundancy, and pseudo-anonymity features have made public blockchain platforms attractive for adoption astechnology platforms for cryptocurrencies . We propose a heuristics-based approach that adds new features associated with money trails to analyze and find suspicious activities . We find that malicious activities (such as Gambling, Phishing, andMoney Laundering) have different cyclic patterns in Ethereum .…

Supercomputing Enabled Deployable Analytics for Disaster Response

MIT SuperCloud precompute range of analytics as files that can be used with standard preinstalled software that does not require network access or additional software . In response to the COVID-19 pandemic, this approach was tested for providing geo-spatial census data to allow quick analysis of demographic data for better responding to emergencies .…

Distinct Angle Problems and Variants

The Erd\H{o}s distinct distance problem is a ubiquitous problem in discrete geometry . We provide upper and lower bounds on a broad class of distinct angle problems . We introduce a new class of asymptotically optimal point configurations with no four cocircular points .…

Visual and Language Navigation A Survey and Taxonomy

Visual-and-Language Navigation is a multi-disciplinary field of increasing importance and with extraordinary practicality . This taxonomy enables researchers to better grasp the keypoint of a specific task and identify directions for future research . For multi-turn tasks, we divided them into imperative task and interactive task based on whether the agent responsesto the instructions .…