Transfer Learning for Human & AI

We propose a novel clock skew compensation algorithm based on Bresenham’s line drawing algorithm . The proposed algorithm can avoid the effect of limitedfloating-point precision (e.g., 32-bit single precision) on clock skewcompensation .…

Authors propose a news recommendation framework consisting of collaborativenews encoding (CNE) and structural user encoding (SUE) to enhance news and userrepresentation learning . CNE equipped with bidirectional LSTMs encodes newstitle and content collaboratively with cross-selection and cross-attention modules to learn semantic-interactive news representations .…

Students, computational agents and regular software engineers can prove interesting theorems about programs using browser-based, online games . Within an hour, players areable to specify and verify properties of programs that are beyond the capabilities of fully-automated tools . The hour limit includes the time for setting up the system, completing a short tutorial explaining game play and reasoning about simple imperative programs .…

Disease-aware image editing by means of generative adversarial networks constitutes a promising avenue for advancing the use of AI in the healthcare sector . This work was presented in the MedicalImaging meets Neurips Workshop 2020, which was held as part of the 34thConference on Neural Information Processing Systems (NeurIPS 2020) in Vancouver, Canada .…

Domain adaptation seeks to mitigate shift between training on the source and testing on the target domain . Source-free methods replace the source data with a source model by fine-tuning it on target . However, target accuracy is the goal, and so we argue foroptimizing as much as possible on target data .…

Realtime face identification (FID) from a video feed is highlycomputation-intensive, and may exhaust computation resources if performed on adevice with a limited amount of resources (e.g., a mobile device) In general,FID performs better when images are sampled at a higher rate, minimizing falsenegatives .…

ConQX is a method for semantic expansion of spoken queries . It uses the text generation ability of an auto-regressive language model, GPT-2 . To avoid off-topic text generation, we condition the input query to a structured context . We then apply zero-shot, one-shot and few-shot learning to fine-tune BERT and RoBERTa for intent detection .…

This is the first attempt in the field of recommender systems, we believe the insight of this paper is inspirational to future self-supervised learning research based on multi-channel information . We propose a cross-channelmatching representation model based on attentive interaction .…

Trace clustering has been extensively used to preprocess event logs . Bygrouping similar behavior, these techniques guide the identification ofsub-logs, producing more understandable models and conformance analytics . Little attention has been posed to the relationship between eventlog properties and clustering quality .…

In the last few years, the notion of symmetry has provided a powerful lens to view several optimization or sampling problems that arise in theoretical computer science, statistics, machine learning and machine learning . We present two examples of nonconvex problems in optimization and sampling where continuous symmetries play a key role in the development of efficient algorithms .…

We study the computational complexity of two hard problems on determinantalpoint processes (DPPs) One is maximum a posteriori (MAP) inference, i.e., to find a principal submatrix having the maximum determinant . The other isprobabilisticistic inference on exponentiated DPPs (E-DPPs), which can sharpen orweaken the diversity preference .…

Visual perception is still inthe spotlight as a challenging field because the vision sensor has someproblems in obtaining correct scale information with a monocular camera . UWB sensor fusion is proposed in the visual inertial odometry algorithmas a solution to mitigate this problem .…

APIzator is grounded by four common patterns that we extracted by studying real APIzations in GitHub . It is based on a static analysis algorithm that automatically extracts the method parameters and return statements . For 113 (56.50 %) and 115 (57.50%) APIzators extracted identical parameters or return statements, respectively.…

Mobile Edge Learning (MEL) is a collaborative learning paradigm that features distributed training of Machine Learning (ML) models over edge devices (e.g.,IoT devices) In MEL, possible coexistence of multiple learning tasks may arise . Multiobjective optimization problem is formulated to minimize the total energyconsumption and maximize the learning tasks’ accuracy .…

Problem requires boat detection, matching and estimating its GPS location from imagery with no common features . Solution uses recent ML algorithms, the camera-scenegeometry and Bayesian filtering . The proposed pipeline first detects and tracksthe target boat’s location in the image with the strategy of tracking bydetection .…

Sequence-to-sequence learning with neural networks has become the de factostandard for sequence prediction tasks . We develop a neuralparameterization of the grammar which enables parameter sharing over thecombinatorial space of derivation rules without the need for manual featureengineering . We apply this latent neural grammar to various domains — adiagnostic language navigation task designed to test for compositionalgeneralization (SCAN), style transfer, and small-scale machine translation –and find that it performs respectably compared to standard baselines .…

Transfer learning has been proposed to overcome thesedifficulties even across languages, e.g., German ASR trained from an English model . To simplify the transfer, we propose to use an intermediatealphabet, Czech without accents, and document that it is a highly effectivestrategy .…

Autonomous marine vehicles play an essential role in many ocean science and engineering applications . Planning time and energy optimal paths for these vehicles to navigate in stochastic dynamic ocean environments is essential to reduce operational costs . Building a realistic model and solving the modeled MDPbecomes computationally expensive in large-scale real-time applications .…

Developing automatic Math Word Problem (MWP) solvers has been an interest of NLP researchers since the 1960s . InMWPToolkit, we decompose the procedure of existing MWP solvers into multiplecore components and decouple their models into highly reusable modules . We alsoprovide a hyper-parameter search function to boost the performance .…

More predictable words are easier to process – they are read faster and . smaller neural signals associated with processing difficulty, mostnotably, the N400 component of the event-related brain potential . Study finds that the predictions of three top-of-the-line . contemporary language models match the .…

We introduce concurrentmatching logic (CML) to reason about fault-free partial correctness of shared-memory concurrent programs . We also present a soundness proof for CML in terms of operational semantics . Undercertain assumptions, the assertion of CSL can be transformed into the assertionof CML .…

The Moser-Veselov equation arises in the discretization of the Euler-Arnold differential equations governing the motion of a generalized rigid body . We propose three iterative methods for solving the Moser Veselov equations . The first onesplits the orthogonal constraints using the Bregman method .…

Software architectural changes involve more than one module or component and are complex to analyze compared to local code changes . A semanticslice consists of a description of relational information of involved modules, their classes, methods and connected modules in a change instance .…

Structure from motion (SfM) enables us to reconstruct a scene via casual capture from cameras at different viewpoints, and novel view synthesis (NVS) allows us to render a captured scene from a new viewpoint . Both are hard with casual capture and dynamic scenes: SfM produces noisy and spatio-temporallysparse reconstructed point clouds, resulting in NVS with spatio temporallyinconsistent effects .…

The inverse linear-quadratic optimal control problem is a systemidentification problem whose aim is to recover the quadratic cost function and the closed-loop system matrices based on observations of optimaltrajectories . In this paper, the discrete-time, finite-horizon case is considered, where the agents are also assumed to be indistinguishable .…

Accurate and globally referenced global navigation satellite system (GNSS)based vehicular positioning can be achieved in outlier-free open areas . Performance of GNSS can be significantly degraded by outlier measurements, such as multipath effects and non-line-of-sight (NLOS) receptionsarising from signal reflections of buildings .…

Sequence-to-sequence learning with neural networks has become the de factostandard for sequence prediction tasks . We develop a neuralparameterization of the grammar which enables parameter sharing over thecombinatorial space of derivation rules without the need for manual featureengineering . We apply this latent neural grammar to various domains — adiagnostic language navigation task designed to test for compositionalgeneralization (SCAN), style transfer, and small-scale machine translation –and find that it performs respectably compared to standard baselines .…

Explainable Recommendation has attracted a lot of attention due to a renewed interest in explainable artificial intelligence . This paper aims to show how the characteristics of a classicalpost-hoc model based on surrogates is strongly model-dependent and does not prove to be accountable for the explanations generated .…

Previous studies have shown that up to 99.59 % of the Java appsusing crypto APIs misuse the API at least once . However, these studies have been conducted on Java and C, while empirical studies for other languages aremissing . With this analysis, we analyzed 895 popular Python projects from GitHub and 51MicroPython projects for embedded devices .…

Target-oriented opinion words extraction (TOWE) is a newsubtask of target-oriented sentiment analysis that aims to extract opinionwords for a given aspect in text . Current state-of-the-art methods leverageposition embeddings to capture the relative position of a word to the target .…

VORRT-COLREGs combines velocityobstacles (VO) and rapidly-exploring random trees (RRT) to generate safetrajectories for autonomous surface vessels . RRT generates a set of way points and the velocity obstaclesmethod ensures safe travel between way points . We also ensure that the actions of ASVs do not violate maritime collision guidelines.…

Graph feature extraction is a fundamental task in graphs analytics . Using feature vectors (graph descriptors) in tandem with data mining algorithms thatoperate on Euclidean data, one can solve problems such as classification,clustering, and anomaly detection on graph-structured data . In this paper, we present single-pass streaming algorithms to approximate structural features of graphs (counts of subgraphs of order $k \geq4$) Operating on edge streams allows us to avoid keeping the entire graph in memory, and controlling the sample size enables us to control the time taken by the algorithm .…

This work considers the problem of learning cooperative policies in multi-agent settings with partially observable and non-stationary environments without a communication channel . We propose two novel ways of integrating information across agents and time in MACRPO . The code is available online athttps://://://github.com/kargarisaac/macrpo.…

Real-time optimization with persistent parameter adaptation (ROPA) avoids waiting for a steady-state before triggering the optimization cycle . ROPA’s performance is similar to dynamic RTO’s performance with a much lower computational cost, authors say . They say ROPA can be scheduled at any desired rate and is easier to implement than traditional RTO .…

Designing and implementing efficient parallel priority schedulers is an active research area . The Stealing Multi-Queue is a cache-efficient variant of the multi-queue, which leverages both queueaffinity and queue-affinity . We provide ageneral SMQ implementation that can surpass state-of-the-art scheduler such asGalois and PMOD in terms of performance on graph-processing benchmarks .…

Independent deeply learned matrix analysis (IDLMA) is one of the state-of-the-art multichannel audio source separation methods . We focus on a blind source separation counterpart of IDLMA, independent low-rank matrix analysis . It uses nonnegative matrixfactorization (NMF) as the source model, which can capture source spectral components that only appear in the target mixture .…

Byzantine consensus is a classical problem in distributed computing . Each node in a synchronous system starts with a binary input . The goal is to reach agreement in the presence of Byzantine faulty nodes . We consider the setting where communication between nodes is modelled via a directed hypergraph .…

We study learning dynamics induced by strategic agents who repeatedly play agame with an unknown payoff-relevant parameter . In this dynamics, a beliefestimate of the parameter is repeatedly updated given players’ strategies and payoffs using Bayes’s rule . We show that beliefs and strategies converge to a fixed point, where thebelief consistently estimates the payoff distribution for the strategy, and the strategy is an equilibrium corresponding to the belief .…

We introduce MULTI-EURLEX, a new multilingual dataset for topic classification of legal documents . The dataset comprises 65k European Union laws, officially translated in 23 languages, annotated with multiple labels from the EUROVOC taxonomy . We use thedataset as a testbed for zero-shot cross-lingual transfer, where we exploitannotated training documents in one language (source) to classify documents in another language (target) We find that fine-tuning a multilingually pretrainedmodel in a single source language leads to catastrophicforgetting of multilingual knowledge and, consequently, poorzero-shot transferto other languages .…

Humans need easy and effective ways of specifying tasks to robots . Goalimages are one popular form of task specification, as they are already grounded in the robot’s observation space . Natural language provides aconvenient and flexible alternative for task specification .…

Efficiently exploiting servers in data centers requires performance analysismethods that account not only for the stochastic nature of demand but also forserver heterogeneity . New analytical methods that exploit its product-form stationary distribution to understand the joint impact of the speeds and buffer lengths of servers on performance.…

We present quadratically convergent algorithms to compute the extremal value of a real parameter for which a given rational transfer function of a lineartime-invariant system is passive . This problem is formulated for bothcontinuous-time and discrete-time systems . Our new methods make use of the HybridExpansion-Contraction algorithm, which we extend and generalize to the setting of root-max problems .…

This paper presents a comprehensive evaluation of instance segmentation models with respect to real-world image corruptions . The out-of-domain image evaluation shows thegeneralization capability of models . The study includes state-of theart networkarchitectures, network backbones, normalization layers, models trained startingfrom scratch or ImageNet pretrained networks, and the effect of multi-tasktraining on robustness and generalization on the models .…

Recommender systems have become a ubiquitous part of modern web applications . Users possess an intuitive andophisticated understanding of the recommender system’s behavior . Users, based upon their understanding, attitude, and intentions, changetheir interactions to evoke desired recommender behavior. We describe this understanding as the Theory of the Recommender.…

Graph Contrastive Learning (GCL) establishes a new paradigm for learninggraph representations without human annotations . The success behind GCL is still left somewhatmysterious, although remarkable progress has been witnessed recently . We develop an easy-to-use library PyGCL, featuring modularized CL components, standardizedevaluation, and experiment management .…

We are the first to use acquired datafrom the near-infrared and ultraviolet spectra for non-photorealistic rendering . Several plant and animal species are more comprehensively understood bymultispectral analysis . Traditional NPR techniques ignore unique information outside the visible spectrum . We introduce algorithms and principles for processing wavelength dependent surface normals and reflectance .…

There is still no technology that can easily turn a staticscan into an animatable avatar . Automating the creation of such avatars wouldenable many applications in games, social networking, animation, and AR/VR toname a few. The key problem is one of representation.…

This paper proposes an improved global navigation satellite system (GNSS)positioning method . Instead of relying on the timedifference carrier phase (TDCP) which only considers two neighboring epochs, this paper proposed to employ the carrier-phase measurements inside a window, the so-called windowcarrier-phase (WCP) to constrain the states inside a factor graph .…

We study fair and efficient allocation of divisible goods, in an onlinemanner, among $n$ agents . The goods arrive online in a sequence of $T$ time periods . The agents’ values for a good are revealed only after its arrival, and the algorithm needs to fractionally allocate the good, immediately andirrevocably, among the agents .…

This paper presents the overall system — design,hardware, and software — for Quori, a novel, affordable, socially interactive robot platform for facilitating non-contact human-robot interaction research . Initial Quori testing and a six-month deployment are presented . Ten Quori platforms have been awarded to a diversegroup of researchers from across the United States to facilitate HRI research to build a community database from a common platform .…