This paper presents a new sequence-to-sequence (seq2seq) pre-training methodPoDA (Pre-training of Denoising Autoencoders), which learns representationssuitable for text generation tasks. Unlike... | Continue reading
Large-scale particle physics experiments face challenging demands forhigh-throughput computing resources both now and in the future. Newheterogeneous computing paradigms on dedicated hardware... | Continue reading
Active learning (AL) on attributed graphs has received increasing attentionwith the prevalence of graph-structured data. Although AL has been widelystudied for alleviating label sparsity issues... | Continue reading
Let $f_n(x_0, x_1, \ldots, x_{n-1})$ denote the algebraic normal form(polynomial form) of a rotation symmetric (RS) Boolean function of degree $d$in $n \geq d$ variables and let $wt(f_n)$ denote... | Continue reading
The Compact Linear Collider (CLIC) is a proposed TeV-scale high-luminosityelectron-positron collider. For an optimal exploitation of its physicspotential, CLIC is foreseen to be built and... | Continue reading
Open-source developers, particularly the elite developers, maintain a diverseportfolio of contributing activities. They do not only commit source code butalso spend a significant amount of... | Continue reading
We study the statistical underpinnings of life. We question some commonassumptions about the thermodynamics of life and illustrate how, contrary towidespread belief, even in a closed system... | Continue reading
Bidirectional Encoder Representations from Transformers (BERT) represents thelatest incarnation of pretrained language models which have recently advanced awide range of natural language... | Continue reading
Monolingual data has been demonstrated to be helpful in improving thetranslation quality of neural machine translation (NMT). The current methodsstay at the usage of word-level knowledge, such... | Continue reading
Election system vendors are marketing ballot-marking devices (BMDs) as auniversal system, and some states are deploying them for all voters, not justthose who need a BMD to vote independently.... | Continue reading
Previous research on empathetic dialogue systems has mostly focused ongenerating responses given certain emotions. However, being empathetic not onlyrequires the ability of generating emotional... | Continue reading
Recently, a higher competition in logistics business introduces newchallenges to the vehicle routing problem (VRP). Re-route planning, also knownas dynamic VRP, is one of the important... | Continue reading
ConArg is a suite of tools that offers a wide series of applications fordealing with argumentation problems. In this work, we present the advances wemade in implementing a ranking-based... | Continue reading
We use an analogy between non-isomorphic mathematical structures defined overthe same set and the algebras induced by associative and causal levels ofinformation in order to argue that... | Continue reading
In this paper, we propose Latent Relation Language Models (LRLMs), a class oflanguage models that parameterizes the joint distribution over the words in adocument and the entities that occur... | Continue reading
Most functional languages rely on some garbage collection for automaticmemory management. They usually eschew reference counting in favor of a tracinggarbage collector, which has less... | Continue reading
Most of the existing generative adversarial networks (GAN) for textgeneration suffer from the instability of reinforcement learning trainingalgorithms such as policy gradient, leading to... | Continue reading
Unsupervised word embedding has benefited a wide spectrum of NLP tasks due toits effectiveness of encoding word semantics in distributed wordrepresentations. However, unsupervised word embedding... | Continue reading
Electronic health record is an important source for clinical researches andapplications, and errors inevitably occur in the data, which could lead tosevere damages to both patients and hospital... | Continue reading
Dialogue systems dealing with multi-domain tasks are highly required. How torecord the state remains a key problem in a task-oriented dialogue system.Normally we use human-defined features as... | Continue reading
The Dependent Object Types (DOT) calculus aims to formalize the Scalaprogramming language with a focus on path-dependent types $-$ types such as$x.a_1\dots a_n.T$ that depend on the runtime... | Continue reading
Adversarial examples highlight model vulnerabilities and are useful forevaluation and interpretation. We define universal adversarial triggers:input-agnostic sequences of tokens that trigger a... | Continue reading
We start with a brief introduction to reinforcement learning (RL), about itssuccessful stories, basics, an example, issues, the ICML 2019 Workshop on RLfor Real Life, how to use it, study... | Continue reading
We introduce a dataset for studying the evolution of words, constructed fromWordNet and the Google Books Ngram Corpus. The dataset tracks the evolution of4,000 synonym sets (synsets), containing... | Continue reading
Knowledge graph embedding models have gained significant attention in AIresearch. Recent works have shown that the inclusion of background knowledge,such as logical rules, can improve the... | Continue reading
Biological evolution provides a creative fount of complex and subtleadaptations, often surprising the scientists who discover them. However,because evolution is an algorithmic process that... | Continue reading
Multi-task learning and self-training are two common ways to improve amachine learning model's performance in settings with limited training data.Drawing heavily on ideas from those two... | Continue reading
This paper introduces a novel orchestration framework, called CFO(COMPUTATION FLOW ORCHESTRATOR), for building, experimenting with, anddeploying interactive NLP (Natural Language Processing) and... | Continue reading
We address one-shot imitation learning, where the goal is to execute apreviously unseen task based on a single demonstration. While there has beenexciting progress in this direction, most of the... | Continue reading
There is an urgent societal need to assess whether autonomous vehicles (AVs)are safe enough. From published quantitative safety and reliability assessmentsof AVs, we know that, given the goal of... | Continue reading
Spam!: that's what Lorrie Faith Cranor and Brian LaMacchia exclaimed in thetitle of a popular call-to-action article that appeared twenty years ago onCommunications of the ACM. And yet, despite... | Continue reading
In recent years, deep neural networks have found success in replicatinghuman-level cognitive skills, yet they suffer from several major obstacles. Onesignificant limitation is the inability to... | Continue reading
his paper presents two novel approaches to solving the classic board gamemastermind, including a variant of simulated annealing (SA) and a technique weterm maximum expected reduction in... | Continue reading
Active learning aims to develop label-efficient algorithms by sampling themost representative queries to be labeled by an oracle. We describe apool-based semi-supervised active learning... | Continue reading
Solving grounded language tasks often requires reasoning about relationshipsbetween objects in the context of a given task. For example, to answer thequestion "What color is the mug on the... | Continue reading
Abductive reasoning is inference to the most plausible explanation. Forexample, if Jenny finds her house in a mess when she returns from work, andremembers that she left a window open, she can... | Continue reading
Previous work on neural noisy channel modeling relied on latent variablemodels that incrementally process the source and target sentence. This makesdecoding decisions based on partial source... | Continue reading
Biomedical Named Entity Recognition (BioNER) is a crucial step for analyzingBiomedical texts, which aims at extracting biomedical named entities from agiven text. Different supervised machine... | Continue reading
GPT-2 and BERT demonstrate the effectiveness of using pre-trained languagemodels (LMs) on various natural language processing tasks. However, LMfine-tuning often suffers from catastrophic... | Continue reading
This paper describes a new approach on optimization of constraintsatisfaction problems (CSPs) by means of substituting sub-CSPs with locallyconsistent regular membership constraints. The purpose... | Continue reading
Recent years have seen exceptional strides in the task of automaticmorphological inflection generation. However, for a long tail of languages thenecessary resources are hard to come by, and... | Continue reading
Parsing sketches via semantic segmentation is attractive but challenging,because (i) free-hand drawings are abstract with large variances in depictingobjects due to different drawing styles and... | Continue reading
Synthesizing images from a given text description involves engaging two typesof information: the content, which includes information explicitly described inthe text (e.g., color, composition,... | Continue reading
For the task of medical image segmentation, fully convolutional network (FCN)based architectures have been extensively used with various modifications. Arising trend in these architectures is to... | Continue reading
Estimating 3D human pose from monocular images demands large amounts of 3Dpose and in-the-wild 2D pose annotated datasets which are costly and requiresophisticated systems to acquire. In this... | Continue reading
Deep learning has seen tremendous success over the past decade in computervision, machine translation, and gameplay. This success rests in crucial wayson gradient-descent optimization and the... | Continue reading
We introduce a model for mobile, multi-agent information transfer thatincreases the communication covertness through a protocol which also increasesthe information transfer delay. Covertness is... | Continue reading
When working with Git, a popular version-control system, email addresses arepart of the metadata for each individual commit. When those commits are pushedto remote hosting services like GitHub,... | Continue reading