In recent years, machine learning has received increased interest both as anacademic research field and as a solution for real-world business problems.However, the deployment of machine learning... | Continue reading
Abstract reasoning and logic inference are difficult problems for neuralnetworks, yet essential to their applicability in highly structured domains. Inthis work we demonstrate that a well known... | Continue reading
We introduce screen gleaning, a TEMPEST attack in which the screen of amobile device is read without a visual line of sight, revealing sensitiveinformation displayed on the phone screen. The... | Continue reading
Gradient descent and backpropagation have enabled neural networks to achieveremarkable results in many real-world applications. Despite ongoing success,training a neural network with gradient... | Continue reading
The OAuth 2.0 protocol is one of the most widely deployedauthorization/single sign-on (SSO) protocols and also serves as the foundationfor the new SSO standard OpenID Connect. Despite the... | Continue reading
Iverson's APL and its descendants (such as J, K and FISh) are examples of thefamily of "rank-polymorphic" programming languages. The principal controlmechanism of such languages is the general... | Continue reading
Most existing datasets for sound event recognition (SER) are relatively smalland/or domain-specific, with the exception of AudioSet, based on a massiveamount of audio tracks from YouTube videos... | Continue reading
Running machine learning analytics over geographically distributed datasetsis a rapidly arising problem in the world of data management policies ensuringprivacy and data security. Visualizing... | Continue reading
Ownership is the concept of tracking aliases and mutations to data, usefulfor both memory safety and system design. The Rust programming languageimplements ownership via the borrow checker, a... | Continue reading
The path toward realizing the potential of seasonal forecasting and itssocioeconomic benefits depends heavily on improving general circulation modelbased dynamical forecasting systems. To... | Continue reading
We examine how the behavior of software developers changes in response toremoving gamification elements from GitHub, an online platform forcollaborative programming and software development. We... | Continue reading
Machine learning has shown growing success in recent years. However, currentmachine learning systems are highly specialized, trained for particularproblems or domains, and typically on a single... | Continue reading
Graph-structured data are an integral part of many application domains,including chemoinformatics, computational biology, neuroimaging, and socialnetwork analysis. Over the last two decades,... | Continue reading
When data is plentiful, the loss achieved by well-trained neural networksscales as a power-law $L \propto N^{-α}$ in the number of networkparameters $N$. This empirical scaling law holds... | Continue reading
Sentiment analysis as a field has come a long way since it was firstintroduced as a task nearly 20 years ago. It has widespread commercialapplications in various domains like marketing, risk... | Continue reading
Code completion is a popular software development tool integrated into allmajor IDEs. Many neural language models have achieved promising results incompletion suggestion prediction on synthetic... | Continue reading
The Appalachian Trail (AT) is a 2193-mile-long hiking trail in the easternUnited States. The trail has many bends and turns at different length scales,which gives it a nontrivial fractal... | Continue reading
We describe a system for large-scale audiovisual translation and dubbing,which translates videos from one language to another. The source language'sspeech content is transcribed to text,... | Continue reading
EfficientNets are a family of state-of-the-art image classification modelsbased on efficiently scaled convolutional neural networks. Currently,EfficientNets can take on the order of days to... | Continue reading
ML models often exhibit unexpectedly poor behavior when they are deployed inreal-world domains. We identify underspecification as a key reason for thesefailures. An ML pipeline is underspecified... | Continue reading
Dask is a distributed task framework which is commonly used by datascientists to parallelize Python code on computing clusters with littleprogramming effort. It uses a sophisticated... | Continue reading
Encrypted database systems provide a great method for protecting sensitivedata in untrusted infrastructures. These systems are built using eitherspecial-purpose cryptographic algorithms that... | Continue reading
The main source of information regarding ancient Mesopotamian history andculture are clay cuneiform tablets. Despite being an invaluable resource, manytablets are fragmented leading to missing... | Continue reading
We consider the estimation of binary election outcomes as martingales andpropose an arbitrage pricing when one continuously updates estimates. We arguethat the estimator needs to be priced as a... | Continue reading
We present a constant and a recursive relation to define a sequence $f_n$such that the floor of $f_n$ is the $n$th prime. Therefore, this constantgenerates the complete sequence of primes. We... | Continue reading
Graph Neural Networks (GNNs) are the predominant technique for learning overgraphs. However, there is relatively little understanding of why GNNs aresuccessful in practice and whether they are... | Continue reading
We review the Montevideo Interpretation of quantum mechanics, which is basedon the use of real clocks to describe physics, using the framework recentlyintroduced by Hoehn, Smith and Lock to... | Continue reading
In this article we address the question whether it is possible to learn thedifferential equations describing the physical properties of a dynamicalsystem, subject to non-conservative forces,... | Continue reading
Matrix engines or units, in different forms and affinities, are becoming areality in modern processors; CPUs and otherwise. The current and dominantalgorithmic approach to Deep Learning merits... | Continue reading
Increasingly, modern Artificial Intelligence (AI) research has become morecomputationally intensive. However, a growing concern is that due to unequalaccess to computing power, only certain... | Continue reading
The classical development of neural networks has primarily focused onlearning mappings between finite-dimensional Euclidean spaces. Recently, thishas been generalized to neural operators that... | Continue reading
Lyrics parody swaps one set of words that accompany a melody with a new setof words, preserving the number of syllables per line and the rhyme scheme.Lyrics parody generation is a challenge for... | Continue reading
PageRank is a Web page ranking technique that has been a fundamentalingredient in the development and success of the Google search engine. Themethod is still one of the many signals that Google... | Continue reading
Probability trees are one of the simplest models of causal generativeprocesses. They possess clean semantics and -- unlike causal Bayesian networks-- they can represent context-specific causal... | Continue reading
Traditionally, mutation testing generates an abundance of small deviations ofa program, called mutants. At industrial systems the scale and size ofFacebook's, doing this is infeasible. We should... | Continue reading
Hittmeir recently presented a deterministic algorithm that provably computesthe prime factorisation of a positive integer $N$ in $N^{2/9+o(1)}$ bitoperations. Prior to this breakthrough, the... | Continue reading
A recursive descent parser is built from a set of mutually-recursivefunctions, where each function directly implements one of the nonterminals of agrammar. A packrat parser uses memoization to... | Continue reading
Many programming problems call for turning geometrical thoughts into code:tables, hierarchical structures, nests of objects, trees, forests, graphs, andso on. Linear text does not do justice to... | Continue reading
It is non-trivial to design engaging and balanced sets of game rules. Modernchess has evolved over centuries, but without a similar recourse to history,the consequences of rule changes to game... | Continue reading
Widely popular transformer-based NLP models such as BERT and Turing-NLG haveenormous capacity trending to billions of parameters. Current execution methodsdemand brute-force resources such as... | Continue reading
A central claim in quantum cryptography is that secrecy can be provedrigorously, based on the assumption that the relevant information-processingsystems obey the laws of quantum physics. This... | Continue reading
Four years ago, an experimental system known as PilotNet became the firstNVIDIA system to steer an autonomous car along a roadway. This systemrepresents a departure from the classical approach... | Continue reading
In this work, we present an interaction-based approach to learn semanticallyrich representations for the task of slicing vegetables. Unlike previousapproaches, we focus on object-centric... | Continue reading
Philosophers of science have long postulated how collective scientificknowledge grows. Empirical validation has been challenging due to limitationsin collecting and systematizing large... | Continue reading
At the heart of what drives the bulk of innovation and activity in SiliconValley and elsewhere is scalability. This unwavering commitment to scalability-- to identify strategies for efficient... | Continue reading
Deep neural networks require large training sets but suffer from highcomputational cost and long training times. Training on much smaller trainingsets while maintaining nearly the same accuracy... | Continue reading
This paper discusses a novel approach for detecting moving massive objectsbased on the time variation that these objects produce in the localgravitational field measured by several detectors.... | Continue reading
The era of edge computing has arrived. Although the Internet is the backboneof edge computing, its true value lies at the intersection of gathering datafrom sensors and extracting meaningful... | Continue reading