Transfer Learning for Human & AI

We present a symbolic algorithmic approach that allows to compute invariantmanifolds and corresponding reduced systems for differential equations modelingbiological networks . Multiple time scales of a given network are obtained by scaling, based on tropical geometry . Theexistence of invariant manifolds is subject to hyperbolicity conditions, whichwe test algorithmically using Hurwitz criteria .…

Contour integration schemes are a valuable tool for the solution of difficult interior eigenvalue problems . The solution of many large linearsystems with multiple right hand sides may prove a prohibitive computational cost . The number of right-hand sides, and thus, computational cost may be reduced if the projected subspace is created using multiple moments .…

We present a high-order entropy stable discontinuous Galerkin (ESDG) method for nonlinear conservation laws . These methods utilizetreatments of multi-dimensional interfaces and network junctions retain entropy stability when coupling together entropy stable discretizations . Numerical experiments verify the stability of the proposed schemes, and .comparisons…

Many applications are based on the use of efficient Fourier algorithms such as the fast Fourier transform . In a high-dimensional setting it is also already possible to deal with specialsampling sets such as sparse grids or rank-1 lattices . In this paper we proposefast algorithms for high .dimensional…

The problem of data exchange between multiple nodes with (not necessarily uniform) storage and communication capabilities models several current communication problems like Coded Caching, Data shuffling, CodedComputing, etc. The expression of the converse depends only on the number of bits to be moved between differentsubsets of nodes, and does not assume anything further specific about the parameters in the problem .…

Single-task UAVs are increasingly being employed to carry out surveillance,parcel delivery, communication support, and other specific tasks . When thegeographical area of operation of single-task missions is common, e.g., inpost-disaster recovery scenarios, it is more efficient to have multiple tasks carried out as part of a single UAV mission .…

We present a scheme for translating logic programs into algebraic expressions that denote bag relations overground terms of the Herbrand universe . We develop an operational semantics based on term rewriting of thealgebraic expressions . This approach can exploit arithmetic identities and can exploit a range of useful strategies, including lazy strategies that defer work until it becomes possible or necessary .…

Existingnon-contact breath assessments achieve satisfactory results under restricted conditions when human body stays relatively still . When someone moves forward, breath signals detected by depth camera are hidden within signals of trunkdisplacement and deformation . The proposed approach outperforms the comparative methods with the accuracy, precision, recall and F1 of 75.5%, 76.2%, 75.0% and 75.2% .…

Attention modules, as simple and effective tools, have enabled deepneural networks to achieve state-of-the-art results in many domains . Most current models use deterministicattention modules due to their simplicity and ease of optimization . Stochastic counterparts, on the other hand, are less popular despite their potential benefits .…

Categories with feedback, originally proposed by Katis, Sabadini and Walters, are a weakening of the notion of traced monoidal categories . The construction of the free such categories has appeared in several different contexts, and can be considered as statebootstrapping .…

CoRT is a framework andneural first-stage ranking model that leverages contextual representations fromtransformer-based language models to complement candidates from term-based ranking functions while causing no significant delay . CoRTa representation-focused retrieval at web-scale can be realized with latenciesas low as BM25.…

This paper presents a method for local motion planning in unstructuredenvironments with static and moving obstacles such as humans . Given areference path and speed, our optimization-based receding-horizon approach computes a local trajectory that minimizes the tracking error while avoidingobstacles .…

As robots perform manipulation tasks and interact with objects, it isprobable that they accidentally drop objects that subsequently bounce out of their visual fields . To enable robots to recover from such errors, we draw upon the concept ofobject permanence .…

This paper describes how the Koopman operator can be used to generate approximate linear, bilinear, and nonlinear modelrealizations from data . The paper argues in favor of bilinare realizations forcharacterizing systems with unknown dynamics . It argues that every control-affine system admits an infinite-dimensionalbilinearrealization, but does not necessarily admit a linear one .…

SSUSIES contains two methods, speaker separation usingspeaker inventories (SSUSI) and speaker separation using estimated speech(SSUES) SSUSI performs speaker separation with the help of speaker inventory . SSUES is awidely applicable technique that can substantially improve speaker separationperformance using the output of first-pass separation .…

The recent coronavirus pandemic has highlighted the many challenges faced by healthcare, public safety, and economic systems when confronted with asurge in patients that require intensive treatment and a population that must be quarantined or shelter in place . Robotic technologies areinherently programmable, and robotic systems have been adapted and deployed, tosome extent, in the current crisis for such purposes as transport, logistics,and disinfection .…

New navigation system imitates insects using polarized skylight for navigation . However, current PSNS is difficult to obtain three-dimensional (3D) attitude in real time only relying on polarizedskylight . PSNS/GNSS/SINS (globalnavigation satellite system/strapdown inertial navigation system) integrated navigation system was constructed .…

Recent years have seen a surge in the number of available frameworks for speech enhancement (SE) and recognition . These frameworks often rely in isolation on either time-domainsignals or time-frequency (TF) representations of speech data . In this study, we investigate the advantages of each set of approaches by examining their impact on speech intelligibility and quality .…

We present a novel multi-modal unspoken punctuation prediction system for the English language . We demonstrate for the first time that by relying exclusively on synthetic data generated using aprosody-aware text-to-speech system, we can outperform a model trained on human audio recordings .…

Time-frequency audio source separation is usually achieved by estimating theshort-time Fourier transform (STFT) magnitude of each source . This loss doesn’t properly account for some perceptual properties of audio . In this paper, we propose to reformulate phase recovery in audiosource separation as a minimization problem involving Bregman divergences .…

With the advent of the upcoming 5G future, billions of F-RAN(Fog-Radio Access Network) nodes will created and used for for the purpose ofEdge Caching . The total amount of memory deployed at the edge is expected to increase 100 times .…

This paper introduces BIRD, the Big Impulse Response Dataset . This opendataset consists of 100,000 multichannel room impulse responses (RIRs) BIRD is the largest open dataset currently available . These RIRs can be used to perform online data augmentation for scenarios that involve two microphones and multiple sound sources .…

Sentiment analysis methods have become popular for investigating human communication, including discussions related to software projects . We find different SE-specific sentiment analysis tools might lead to contradictory results at a fine-grain level, when used ‘off-the-shelf’ Conversely, platform-specific tuning or retraining might be needed to take into account differences in platform conventions, jargon, or document lengths .…

This paper presents a method and an open-source implementation,Bernstein/B\’ezier Optimal Trajectories (BeBOT) for the generation of trajectories for autonomous system operations . The proposed method is based oninfinite dimensional optimal control formulations of trajectory generation problems . By approximating the trajectories using Bernstein polynomials, these problems can be transcribed as nonlinear programming problems, which can then be solved using off-the-shelf solvers .…

The proposed power pooling function outperforms the state-of-the-art linear softmaxpooling on both coarsegrained and fine-grained metrics . It improvesthe event-based F1 score (which evaluates the detection of event onsets andoffsets) by 11.4% and 10.2% relative on the two datasets . While this paperfocuses on sound event detection applications, the proposed method can be applied to other domains in other domains, such as engineering and other domains where it can be used to detect sound events with weak labels that only specify the types of events .…

We present a planner for large-scale (un)labeled object sorting tasks, whichuses two types of manipulation actions: overhead grasping and planar pushing . The grasping action offers completeness guarantee under mild assumptions, andplanar pushing is an acceleration strategy that moves multiple objects at once .…

Understanding individual mobility behavior is critical for modeling urbantransportation . It provides deeper insights on the generative mechanisms of human movements . Emerging data sources such as mobile phone call detailrecords, social media posts, GPS observations, and smart card transactions have been used before .…

We present the first large scale andprincipled study of account repurposing on Twitter . We find 3,500 repurposed accounts in the Twitter Elections Integrity Datasets . 100,000 accounts that have more than 5,000 followers and were active in the first six months of 2020 using Twitter’s 1% real-timeample .…

The theme of human mobility is transversal to multiple fields of study and applications, from ad-hoc networks to smart cities . In this article, we look at the problem from a location-centricperspective . We investigate methods to extract, classify and quantify thesymbolic locations specified in telco trajectories .…

Most of the state-of-the-art automatic music transcription (AMT) models breakdown the main transcription task into sub-tasks such as onset prediction and offset prediction . We explore the effect that spectrogramreconstruction has on our AMT model . Our proposed model consists of two U-nets: the first U-net transcribes the spectrogram into a posteriorgram, and the secondU-net transforms the posteriorgram back into a spectrogram .…

Graph anomaly detection systems aim at identifying suspicious accounts orbehaviors on social networking sites and e-commercial platforms . Detecting anomalous users at an early stage is crucial to minimize financial loss . Eland exploits the mutual enhancement between behavior predictionand graph anomaly detection .…

Keyword Spotting (KWS) plays a vital role in human-computer interaction for smart on-device terminals and service robots . It remains challenging to achieve the trade-off between small footprint and high accuracy for KWS task . We propose a multi-branch temporalconvolution module (MTConv) that can be trained for better performance .…

Federated learning (FL) is an emerging paradigm for facilitating multiple organizations’ data collaboration without revealing their private data to eachother . Recently, vertical FL, where the participating organizations hold the same set of samples but with disjoint features and only one organization ownsthe labels, has received increased attention .…

We analyse a Singapore-based COVID-19 Telegram group with more than 10,000 participants . Engagement peaked when the Ministry of Health raised the diseasealert level . We find that government-identified misinformation is rare in the group . Messages discussing misinformation express skepticism, we find that messages discussing these pieces of misinformation are often skeptical of each other .…

Large, multi-dimensional spatio-temporal datasets are omnipresent in modern science and engineering . An effective framework for handling such data are Gaussian process deep generative models (GP-DGMs) The SGP-VAE is evaluated in a variety of experiments where it outperforms alternative approaches including multi-output GPs and structuredVAEs .…

Conversational AI systems are being used in personal devices, providing users with highly personalized content . Personalized knowledge graphs are one of the recently proposed methods to store users’ information in a structured form and tailor answers to their liking .…

When autonomous agents interact in the same environment, they must oftencooperate to achieve their goals . One way for agents to cooperate effectively is to form a team, make a binding agreement on a joint plan, and execute it . When agents are self-interested, gains from team formation must be allocated appropriately to incentivize agreement .…

Contextual word embedding models, such as BioBERT and Bio_ClinicalBERT, have achieved state-of-the-art results in biomedical natural language processing . However, such models do not take into consideration expert domain knowledge . UmlsBERT can encode clinical domain knowledge into word embeddings and outperform existing domain-specific models on common named-entityrecognition (NER) and clinical natural language inference clinical NLP tasks .…

Popular task-oriented dialog data sets such as MultiWOZ are created by providing crowd-sourced workers a goal instruction,expressed in natural language, that describes the task to be accomplished . We demonstrate that by using the simulated data, we achieve significant improvements in both low-resource setting as well as in over-all task performance .…

In this paper, we first theoretically show that robust imitation learning can be achieved by optimizing a classification risk with a symmetric loss . We then propose a newimitation learning method that optimizes the classification risk by effectively combining pseudo-labeling with co-training .…

Mining frequent episodes aims at recovering sequential patterns from temporaldata sequences, which can then be used to predict the occurrence of related events in advance . Gradual patterns captureco-variation of complex attributes in the form of ” when X increases/decreases, Y increases/increase” Such patterns may add knowledge to certain applications, such as e-commerce .…

Graph convolution is a fundamental building block for many deep neuralnetworks on graph-structured data . In this paper, we introduce a simple, yetvery effective graph convolutional network with skip connections forsemi-supervised anomaly detection . The effectiveness ofour model is demonstrated through extensive experiments on five benchmarkdatasets, achieving better or comparable anomaly detection results against strong baseline methods .…

Machine learning (ML) sub-disciplines make mathematical assumptions to facilitate proof-writing . But they also limit the applicability of these algorithms to different problemsettings . I argue such assumptions are equally as important in areas of ML with clear social impact, such as fairness .…

Provenance is a record that describes how entities, activities, and agents influenced a piece of data . Graph kernels, on the other hand, have been consistently and successfully used to efficiently classify graphs . In thispaper, we introduce a novel graph kernel called \emph{provenance kernel .…

Distributed Constraint Optimization Problems (DCOPs) are a widely studied framework for coordinating interactions in cooperative multi-agent systems . In classical DCOPs, variables owned by agents are assumed to be discrete . In many applications, continuous-valued variables are more suitable than discrete ones .…

Smart Grids of collaborative netted radars accelerate kill chains through more efficient cross-cueing over centralized command and control . We propose two novel reward-based learning approaches to decentralizednetted radar coordination based on black-box optimization and ReinforcementLearning . To make the RL approach tractable, we use a simplification of the problem that we proved to be equivalent to the initial formulation .…

We consider the fundamental problem of how to automatically construct summarystatistics for implicit generative models . The idea is to frame the task of constructing sufficient statisticsas learning mutual information maximizing representation of the data . Thisrepresentation is computed by a deep neural network trained by a jointstatistic-posterior learning strategy .…

A robust 3D object tracker which continuously tracks surrounding objects and estimates their trajectories is key for self-driving vehicles . Most existing tracking methods employ a tracking-by-detection strategy, which usually involves complex pair-wise similarity computation and neglects the nature of continuous object motion .…

Procedures are an important knowledge component of documents that can beleveraged by cognitive assistants for automation, question-answering or driving a conversation . It is a challenging problem to parse big dense documents likeproduct manuals, user guides to automatically understand which parts are talking about procedures and subsequently extract them .…

A new encoding heuristic encodes the shape of each character through images depicting the letters when printed . Empirically, our method improves models’ robustness against substandard inputs, even in the test scenario where the models are tested with the noises that are beyond what is available during the training phase .…