We are surprised to find that BERT's peak performance of 77% on the ArgumentReasoning Comprehension Task reaches just three points below the averageuntrained human baseline. However, we show... | Continue reading
Many organizations wish to collaboratively train machine learning models ontheir combined datasets for a common benefit (e.g., better medical research, orfraud detection). However, they often... | Continue reading
Deep learning techniques have become the method of choice for researchersworking on algorithmic aspects of recommender systems. With the stronglyincreased interest in machine learning in... | Continue reading
In this paper, we present RayTracer.jl, a renderer in Julia that is fullydifferentiable using source-to-source Automatic Differentiation (AD). Thismeans that RayTracer not only renders 2D images... | Continue reading
We discuss milestones on the tour towards DPLL(MAPF), a multi-agent pathfinding (MAPF) solver fully integrated with the Davis-Putnam-Logemann-Loveland(DPLL) propositional satisfiability testing... | Continue reading
Competency Questions (CQs) for an ontology and similar artefacts aim toprovide insights into the contents of an ontology and to demarcate its scope.The absence of a controlled natural language,... | Continue reading
Humanitarian disasters have been on the rise in recent years due to theeffects of climate change and socio-political situations such as the refugeecrisis. Technology can be used to best mobilize... | Continue reading
We introduce natural adversarial examples -- real-world, unmodified, andnaturally occurring examples that cause classifier accuracy to significantlydegrade. We curate 7,500 natural adversarial... | Continue reading
The proliferation of smart home Internet of Things (IoT) devices presentsunprecedented challenges for preserving privacy within the home. In this paper,we demonstrate that a passive network... | Continue reading
We present a new and independent determination of the local value of theHubble constant based on a calibration of the Tip of the Red Giant Branch(TRGB) applied to Type Ia supernovae (SNeIa). We... | Continue reading
The POSIX shell is a widely deployed, powerful tool for managing computersystems. The shell is the expert's control panel, a necessary tool forconfiguring, compiling, installing, maintaining,... | Continue reading
We show that the imperceptibility of several existing linguisticsteganographic systems (Fang et al., 2017; Yang et al., 2018) relies onimplicit assumptions on statistical behaviors of fluent... | Continue reading
Simulating the flow of different fluids can be a highly computationalintensive process, which requires large amounts of resources. Recently therehas been a lot of research effort directed... | Continue reading
Principles of cognitive economy would require that concepts about objects,properties and relations should be introduced only if they simplify theconceptualisation of a domain. Unexpectedly,... | Continue reading
Deep learning based question answering (QA) on English documents has achievedsuccess because there is a large amount of English training examples. However,for most languages, training examples... | Continue reading
Games have benchmarked AI methods since the inception of the field, withclassic board games such as Chess and Go recently leaving room for video gameswith related yet different sets of... | Continue reading
Graphs are an essential part of many machine learning problems such asanalysis of parse trees, social networks, knowledge graphs, transportationsystems, and molecular structures. Applying... | Continue reading
Recent research has introduced ideas from concept drift into process miningto enable the analysis of changes in business processes over time. This streamof research, however, has not yet... | Continue reading
We present an improved determination of the Hubble constant (H0) from HubbleSpace Telescope (HST) observations of 70 long-period Cepheids in the LargeMagellanic Cloud. These were obtained with... | Continue reading
We provide here a general mathematical framework to model attitudes towardsambiguity which uses the formalism of quantum theory as a "purely mathematicalformalism, detached from any physical... | Continue reading
Our game Pommerman is based on the console game Bommerman. The game starts onan 11 by 11 platform. Pommerman is a multi-agent environment and is made up ofa set of different situations and... | Continue reading
Feed-forward neural networks can be understood as a combination of anintermediate representation and a linear hypothesis. While most previous worksaim to diversify the representations, we... | Continue reading
We can define a neural network that can learn to recognize objects in lessthan 100 lines of code. However, after training, it is characterized bymillions of weights that contain the knowledge... | Continue reading
Filters of convolutional networks used in computer vision are oftenvisualized as image patches that maximize the response of the filter. We usethe same approach to interpret weight matrices in... | Continue reading
Personalized news recommendation is very important for online news platformsto help users find interested news and improve user experience. News and userrepresentation learning is critical for... | Continue reading
Derivatives play a critical role in computational statistics, examples beingBayesian inference using Hamiltonian Monte Carlo sampling and the training ofneural networks. Automatic... | Continue reading
Coreference resolution is a key problem in natural language understandingthat still escapes reliable solutions. One fundamental difficulty has been thatof resolving instances involving pronouns... | Continue reading
We propose Stable Yet Memory Bounded Open-Loop (SYMBOL) planning, a generalmemory bounded approach to partially observable open-loop planning. SYMBOLmaintains an adaptive stack of Thompson... | Continue reading
Airspace models have played an important role in the development andevaluation of aircraft collision avoidance systems for both manned and unmannedaircraft. As Urban Air Mobility (UAM) systems... | Continue reading
Data exfiltration attacks have led to huge data breaches. Recently, theEquifax attack affected 147M users and a third-party library - Apache Struts -was alleged to be responsible for it. These... | Continue reading
We introduce the concept of design continuums for the data layout ofkey-value stores. A design continuum unifies major distinct data structuredesigns under the same model. The critical insight... | Continue reading
An important step in the development of value alignment (VA) systems in AI isunderstanding how values can interrelate with facts. Designers of future VAsystems will need to utilize a hybrid... | Continue reading
We introduce a new generative model where samples are produced via Langevindynamics using gradients of the data distribution estimated with scorematching. Because gradients might be ill-defined... | Continue reading
Topic modeling analyzes documents to learn meaningful patterns of words.Dynamic topic models capture how these patterns vary over time for a set ofdocuments that were collected over a large time... | Continue reading
In this paper we propose a novel neural approach for automatic deciphermentof lost languages. To compensate for the lack of strong supervision signal, ourmodel design is informed by patterns in... | Continue reading
The ability to forecast a set of likely yet diverse possible future behaviorsof an agent (e.g., future trajectories of a pedestrian) is essential forsafety-critical perception systems (e.g.,... | Continue reading
Handwritten text recognition is challenging because of the virtually infiniteways a human can write the same message. Our fully convolutional handwritingmodel takes in a handwriting sample of... | Continue reading
We quickly approach a "pervasive future" where pervasive computing is thenorm. In this scenario, humans are surrounded by a multitude of heterogeneousdevices that assist them in almost every... | Continue reading
Organic updates (from a member's network) and sponsored updates (or ads, fromadvertisers) together form the newsfeed on LinkedIn. The newsfeed, the defaulthomepage for members, attracts them to... | Continue reading
Malware attacks represent a significant part of today's security threats.Software guard extensions (SGX) are a set of hardware instructions introducedby Intel in their recent lines of processors... | Continue reading
Bitcoin is a peer-to-peer payment system proposed by Nakamoto in 2008.Properties of the bitcoin backbone protocol have been investigated in somedepth: the blockchain growth property quantifies... | Continue reading
Local differential privacy (LDP) has been deemed as the de facto measure forprivacy-preserving distributed data collection and analysis. Recently,researchers have extended LDP to the basic data... | Continue reading
Audio steganography is a collection of techniques for concealing theexistence of information by embedding it within a non-secret audio, which isreferred to as carrier. Distinct from... | Continue reading
As spoken dialogue systems and chatbots are gaining more widespread adoption,commercial and open-sourced services for natural language understanding areemerging. In this paper, we explain how we... | Continue reading
End-to-end Task-oriented Dialogue Systems (TDSs) have attracted a lot ofattention for their superiority (e.g., in terms of global optimization) overpipeline modularized TDSs. Previous studies on... | Continue reading
Semantic parsing converts natural language queries into structured logicalforms. The paucity of annotated training samples is a fundamental challenge inthis field. In this work, we develop a... | Continue reading
The introduction of pre-trained language models has revolutionized naturallanguage research communities. However, researchers still know relativelylittle regarding their theoretical and... | Continue reading