Transfer Learning for Human & AI

BeamTransformer seeks to optimize modeling of sequential sequentialrelationship among signals from different spatial direction . The results indicate that a successful incorporation of microphonearray signals can lead to remarkable gains . Beam transformer takes onestep further, as speech from overlapped speakers have been internally separated internally separated into different beams.…

A common problem in the task of human-object interaction (HOI) detection is that numerous HOI classes have only a small number of labeled examples . The lack ofpositive labels can lead to low classification accuracy for these classes . In this paper, we model the correlations as action co-occurrence matrices and present techniques to learn these priors and leverage them for more effective training, especially onrare classes .…

Neural relation extraction models have shown promising results in recent years . However, the model performance drops dramatically given only a few training samples . We propose a framework considering bothlabel-agnostic and label-aware semantic mapping information for low resource extraction .…

The game is designed to elicit status-driven leader-follower behaviours as a meansto observe and influence social hierarchy . It involves a Bach/Stravinsky game of deference in a group, in which people on each turn can either invest with another player or hope someone invests with them .…

Multilingual neural machine translation (MNMT) learns to translate multiple language pairs with a single model . The heavy data imbalance between languages hinders the model from performing uniformly across language pairs . We propose a new learning objective for MNMT based on distributionally robust optimization .…

A large class of decision making under uncertainty problems can be describedvia Markov decision processes (MDPs) or partially observable MDPs . Traditionally, policy synthesis techniques are proposed such that atotal expected cost or reward is minimized or maximized . However, optimality inthe total expected cost sense is only reasonable if system behavior in thelarge number of runs is of interest, which has limited the use of such policies in practical mission-critical scenarios .…

In this paper, we present a derivative-based, functional recognizer and parsing generator for visibly pushdown grammars . We also present a translator that converts a tagged CFG to a pushdown grammar in a sound way . The performance of the parser is compared with a popular parsing tool ANTLR and other popular hand-crafted parsers .…

We present an algorithm for strongly refuting smoothed instances of allBoolean CSPs . The smoothed model is a hybrid between worst and average-case input models . For an $n-variable smoothed instance of a $k$-arity CSP, our algorithm runs in $n^{O(\ell) time .…

Biometric systems are vulnerable to Presentation Attacks (PA) performedusing various Presentation Attack Instruments (PAIs) There arenumerous PAD techniques based on both deeplearning and hand-crafted features . The proposed method, denoted as IF-OM, is based on a global-local view coupled with De-Folding and De-Mixing to derive the task-specific representation for PAD .…

The multiple traveling salesman problem (mTSP) is a well-known NP-hard problem with numerous real-world applications . The objective is to minimize the max tour length(sum of Euclidean distances) among all agents . In DAN, agents learn fully decentralizedpolicies to collaboratively construct a tour, by predicting the futuredecisions of other agents .…

We consider the online search problem in which a server starting at the origin of a $d$-dimensional Euclidean space has to find an arbitraryhyperplane . The best-possible competitive ratio and the length of the shortestcurve from which each point on the $d#-dimensional unit sphere can be seen are within a constant factor of each other .…

In this paper, we study a hybrid visualization, which combines aspects ofword clouds and point labeling . Point feature labeling is a classical problem in cartography and GIS that has been extensively studied for geospatial point data . We show that computing a minimum set of such rectangles is NP-hard .…

Given the difficulty of reproducing machinelearning (ML) experiments, there have been significant efforts in reducing thevariance of these results . As in any science, the ability to consistentlyreproduce results effectively strengthens the underlying hypothesis of thework . The contribution of this work is a framework that is able tore tore toreproduce consistent results and provides a means of easily creating, training,and evaluating natural language processing (NLP) deep learning (DL) models .…

We present a novel method to perform numerical integration over curvedpolyhedra enclosed by high-order parametric surfaces . We show that folded cells do not cause any issues in practice as they are only used to numericallycompute certain integrals of interest .…

Evolutionary Plastic Recurrent Neural Networks (EPRNN) incorporates with nested loops for metalearning . An outer loop searches for optimal initial parameters of the neuralnetwork and learning rules; an inner loop adapts to specific tasks . In theinner loop of EPRNN, we effectively attain both long term memory and short termmemory by forging plasticity with recursion-based learning mechanisms, both of which are believed to be responsible for memristance in BNNs .…

We present a novel method to perform numerical integration over curvedpolyhedra enclosed by high-order parametric surfaces . We show that folded cells do not cause any issues in practice as they are only used to numericallycompute certain integrals of interest .…

In the oracle identification problem we have oracle access to bits of anunknown string $x$ of length $n$ with the promise that it belongs to a knownset $C\subseteq\{0,1\}^n$. The goal is to identify $x $X$ using as few queries tothe oracle as possible .…

Internet of Things devices have seen a rapid growth and popularity in recent years . Traditional security systems are not able to detect unknown malware as they use signature-based methods . The proposed approach is to faster detect and classify new malware (zero-daymalware) The experiment results show that our method can satisfy the accuracyrequirement of practical application.…

Survey aims to provide a timely and comprehensive overview of the recent trends of deep reinforcement learning in recommender systems . This survey serves as introductory material for readers from academia and industry into the topic . We provide a taxonomy of currentDRL-based recommender system and a summary of existing methods .…

Existing question answering (QA) datasets are created mainly for theapplication of having AI to be able to answer questions asked by humans . But ineducational applications, teachers and parents sometimes may not know what questions they should ask a child that can maximize their language learning results .…

ArchivalQA is a large question answering dataset consisting of 1,067,056 question-answer pairs . The current datasets are essentially designed for synchronic document collections (e.g., Wikipedia) Temporal newscollections such as long-term news archives spanning several decades are rarely used in training the models .…

The widespread use of Artificial Intelligence in consequential domains, such as healthcare and parole decision-making systems, has drawn intensescrutiny on the fairness of these methods . However, ensuring fairness is often insufficient as the rationale for a contentious decision needs to be audited,understood, and defended .…

We present a novel method to perform numerical integration over curvedpolyhedra enclosed by high-order parametric surfaces . We show that folded cells do not cause any issues in practice as they are only used to numericallycompute certain integrals of interest .…

Spoken question answering (SQA) requires fine-grained understanding of both documents and questions for the optimal answer prediction . Wethen propose to learn noise-invariant utterance representations in acontrastive objective by adopting multiple augmentation strategies, includingspan deletion and span substitution . Temporal-Alignmentattention to semantically align the speech-text clues in the learned commonspace and benefit the SQA tasks .…

Graph neural networks have triggered a resurgence of graph-based text classification . We suggest that future studies use atleast an MLP baseline to contextualize the results . We provide recommendations for the design and training of such a baseline . We further fine-tune DistilBERT for comparison and find that itoutperforms all state-of-the-art models .…

Thispaper takes a recent unsupervised method for salience detection derived fromBarthes Cardinal Functions and theories of surprise and applies it to longernarrative forms . We use a novel approach to derive salience annotation using chapter-aligned summaries from the Shmoop corpus for classic literary works .…

Smart factories are equipped with machines that can sense their manufacturing environments, interact with each other, and control production processes . We show how machine learning that is specificallytailored towards industrial applications can help in knowledge graphcompletion . In particular, we show how knowledge completion can benefit from event logs that are common in smart factories .…

RoadAtlas is a novel end-to-endintegrated system that can support 1) road defect detection, 2) road markingparsing, 3) a web-based dashboard for presenting and inputting data by users,and 4) a backend containing a well-structured database and developed APIs . This can effectively address the issue of an expensiveand time-consuming process for professional inspectors to review the street manually .…

Computational models inspired by visual perception have thecharacteristics of complexity and diversity, as they come from many subjectssuch as cognition science, information science, and artificial intelligence . Through this survey, it will provide a comprehensivereference for research in this direction .…

The previous proposed autoregressive formulation for EL suffers from high computational cost due to a complex (deep) decoder . We augment the generative objective with an extra discriminative component, i.e., a correction term which lets us directly optimize thegenerator’s ranking .…

As online shopping prevails and e-commerce platforms emerge, there is atremendous number of parcels being transported every day . The large amount (beyond $10^5$) of daily parcels, thevariability and non-Markovian characteristics of parcel information imposedifficulties on attaining (near-) optimal solution without violating constraints excessively .…

The stochastic approximation (SA) algorithm is a widely used probabilisticmethod for finding a solution to an equation of the form$\mathbf{f}(\boldsymbol{\theta) = \mathbf {R}$ The algorithm is widely used in Reinforcement Learning (RL) In this note, we provide a general theory of convergence for batchasymchronous stochastastic approximation, that works whether the updates use a local clock or a global clock, for the case where the measurement noises form a martingale difference sequence .…

Robotic touch, particularly when using soft optical tactile sensors, suffers from distortion caused by motion-dependent shear . In this work, we propose a supervised convolutionaldeep neural network model that learns to disentangle, in the latent space, the components of sensor deformations caused by contact geometry from those due to movement .…

Predictive Process Monitoring techniques typically construct a predictive model based on past process executions, and then use it to predict the future of new ongoing cases . This can make PredictiveProcess Monitoring too rigid to deal with the variability of processes working in real environments that continuously evolve and/or exhibit new variantbehaviours over time .…

The openness feature of Twitter allows programs to generate and control Twitter accounts automatically via the Twitter API . These accounts, which are known as bots, can automatically perform actions such as tweeting, re-tweeting,following, unfollowing, or direct messaging other accounts .…

In this paper, we propose models that leverage structured knowledge in different components of pre-trained models . Our results show that the proposed models perform better than previous BERT-based state-of-the-art models . Although our models are proposed for NLI, they can be easily extended to other sentence or sentence-pair classification problems .…

TrAISformer-a novel deep learning architecture that canforecast vessel positions using AIS (Automatic Identification System)observations . The model encodes complexmovement patterns in AIS data in high-dimensional vectors, then applies atransformer to extract useful long-term correlations from sequences of thoseembeddings to sample future vessel positions .…

Combining a pretrained language model (PLM) with textual patterns has been shown to help in both zero- and few-shot settings . For zero-shot performance, it makes sense to design patterns that closely resemble the text seen during self-supervised pretraining because the model has never seen anything else .…

In the oracle identification problem we have oracle access to bits of anunknown string $x$ of length $n$ with the promise that it belongs to a knownset $C\subseteq\{0,1\}^n$. The goal is to identify $x $X$ using as few queries tothe oracle as possible .…

We investigate the problem of parameter synthesis for time Petri nets with acost variable that evolves both continuously with time, and discretely when firing transitions . We provide symbolicsemi-algorithms for the two synthesis problems . We also show how to modify them for the case whenparameter values are integers .…

In good-enoughsynthesis, the system is required to generate a satisfying computation only if one exists . Formally, an input sequence x is hopeful if there exists some output sequence y such that the computation x \otimes ysatisfies \psi . In practice, the requirement to satisfy the specification in all environments is often too strong, and it is common to add assumptions on the environment .…

We present a novel method to perform numerical integration over curvedpolyhedra enclosed by high-order parametric surfaces . We show that folded cells do not cause any issues in practice as they are only used to numericallycompute certain integrals of interest .…

The analysis of differential gene expression from RNA-Seq data has become a standard for several research areas mainly involving bioinformatics . This paper presents a review of differential expression analysis pipeline, addressing its steps and the respective objectives, the principal methods available in each step and their properties .…

The analysis of differential gene expression from RNA-Seq data has become a standard for several research areas mainly involving bioinformatics . This paper presents a review of differential expression analysis pipeline, addressing its steps and the respective objectives, the principal methods available in each step and their properties .…

We present a novel method to perform numerical integration over curvedpolyhedra enclosed by high-order parametric surfaces . We show that folded cells do not cause any issues in practice as they are only used to numericallycompute certain integrals of interest .…

Vision-based depth estimation is a key feature in autonomous systems . Camera-only methods suffer from scale-ambiguity and infinite-depth problems . We introduce a novel framework, called LiDARTouch, toestimate dense depth maps from monocular images with the help of “touches” of LiDAR, i.e.,…

MOLD is the firstdataset of its kind compiled for Marathi, thus opening a new domain for research in low-resource Indo-Aryan languages . The widespread presence of offensive language on social media motivated the development of systems capable of recognizing such content automatically .…

The analysis of differential gene expression from RNA-Seq data has become a standard for several research areas mainly involving bioinformatics . This paper presents a review of differential expression analysis pipeline, addressing its steps and the respective objectives, the principal methods available in each step and their properties .…

Recent work on multilingual AMR-to-text generation has exclusively focused on data augmentation strategies that utilize silver AMR . However, this assumes a high quality of generated AMRs, potentially limiting the transferability to the task . In this paper, we investigate different techniques forautomatically generating AMR annotations .…

The Wasserstein barycenter has been widely studied in various fields, including natural language processing, and computer vision . It requires a quadratic timewith respect to the number of supports . By contrast, the tree-Wasserstein distance can be computed in linear time and allows for the fast comparison of a large number of distributions .…