When reverse engineering a binary, the analyst must first understand thesemantics of the binary's functions through either manual or automaticanalysis. Manual semantic analysis is... | Continue reading
We present a selection criterion for the Euclidean metric adapted duringwarmup in a Hamiltonian Monte Carlo sampler that makes it possible for asampler to automatically pick the metric based on... | Continue reading
We introduce a pretraining technique called Selfie, which stands forSELF-supervised Image Embedding. Selfie generalizes the concept of maskedlanguage modeling to continuous data, such as images.... | Continue reading
In this paper, we use a science fiction theme (i.e. the iconic lightsaberfrom the Star Wars universe) as a pedagogical tool to introduce aspects ofnonlinear electrodynamics due to the quantum... | Continue reading
Over-dependence on domain ontology and lack of knowledge sharing acrossdomains are two practical and yet less studied problems of dialogue statetracking. Existing approaches generally fall short... | Continue reading
Graph processing has become an important part of various areas, such asmachine learning, computational sciences, medical applications, social networkanalysis, and many others. Various graphs,... | Continue reading
Modern Haskell supports zero-cost coercions, a mechanism where types thatshare the same run-time representation may be freely converted between. To makesure such conversions are safe and... | Continue reading
In this paper, we introduce the Butterfly Transform (BFT), a light weightchannel fusion method that reduces the computational complexity of point-wiseconvolutions from O(n^2) of conventional... | Continue reading
The capacity of a neural network to absorb information is limited by itsnumber of parameters. Conditional computation, where parts of the network areactive on a per-example basis, has been... | Continue reading
Research in Graph Signal Processing (GSP) aims to develop tools forprocessing data defined on irregular graph domains. In this paper we firstprovide an overview of core ideas in GSP and their... | Continue reading
Recent breakthroughs in AI for multi-agent games like Go, Poker, and Dota,have seen great strides in recent years. Yet none of these games address thereal-life challenge of cooperation in the... | Continue reading
In this paper, we present Ara, a 64-bit vector processor based on the version0.5 draft of RISC-V's vector extension, implemented in GlobalFoundries 22FDXFD-SOI technology. Ara's... | Continue reading
The paper analyzes the accuracy of publicly available object-recognitionsystems on a geographically diverse dataset. This dataset contains householditems and was designed to have a more... | Continue reading
Turi and Plotkin's bialgebraic semantics is an abstract approach tospecifying the operational semantics of a system, by means of a distributivelaw between its syntax (encoded as a monad) and its... | Continue reading
Autonomous multiple tasks learning is a fundamental capability to developversatile artificial agents that can act in complex environments. In real-worldscenarios, tasks may be interrelated (or... | Continue reading
Neural word representations are at the core of many state-of-the-art naturallanguage processing models. A widely used approach is to pre-train, store andlook up word or character embedding... | Continue reading
We propose a new metaheuristic training scheme that combines StochasticGradient Descent (SGD) and Discrete Optimization in an unconventional way. Ouridea is to define a discrete neighborhood of... | Continue reading
We introduce Deep Reasoning Networks (DRNets), an end-to-end framework thatcombines deep learning with reasoning for solving complex tasks, typically inan unsupervised or weakly-supervised... | Continue reading
This paper is a reproduction of work by Ray et al. which claimed to haveuncovered a statistically significant association between eleven programminglanguages and software defects in projects... | Continue reading
Binarized Neural Networks, a recently discovered class of neural networkswith minimal memory requirements and no reliance on multiplication, are afantastic opportunity for the realization of... | Continue reading
This is a graduate-level introduction to C*-algebras, Hilbert C*-modules,vector bundles, and induced representations of groups and C*-algebras, withapplications to quantization theory, phase... | Continue reading
Global acceptance of Emojis suggests a cross-cultural, normative use ofEmojis. Meanwhile, nuances in Emoji use across cultures may also exist due tolinguistic differences in expressing emotions... | Continue reading
In this paper we consider clustering problems in which each point is endowedwith a color. The goal is to cluster the points to minimize the classicalclustering cost but with the additional... | Continue reading
Topology applied to real world data using persistent homology has started tofind applications within machine learning, including deep learning. We presenta differentiable topology layer that... | Continue reading
Low bit-width integer weights and activations are very important forefficient inference, especially with respect to lower power consumption. Wepropose Monte Carlo methods to quantize the weights... | Continue reading
Integrating logical reasoning within deep learning architectures has been amajor goal of modern AI systems. In this paper, we propose a new directiontoward this goal by introducing a... | Continue reading
Deep learning (DL) has solved a problem that as little as five years ago wasthought by many to be intractable - the automatic recognition of patterns indata; and it can do so with accuracy that... | Continue reading
In this paper, we propose a new framework for mitigating biases in machinelearning systems. The problem of the existing mitigation approaches is thatthey are model-oriented in the sense that... | Continue reading
Due to their hunger for big data, modern deep learning models are trained inparallel, often in distributed environments, where communication of modelupdates is the bottleneck. Various update... | Continue reading
Program repair is an important but difficult software engineering problem.One way to achieve a "sweet spot" of low false positive rates, whilemaintaining high enough recall to be usable, is to... | Continue reading
Region proposal mechanisms are essential for existing deep learningapproaches to object detection in images. Although they can generally achieve agood detection performance under normal... | Continue reading
Social Media are nowadays the privileged channel for information spreadingand news checking. Unexpectedly for most of the users, automated accounts, alsoknown as social bots, contribute more and... | Continue reading
Gradient Boosting Machine (GBM) is an extremely powerful supervised learningalgorithm that is widely used in practice. GBM routinely features as a leadingalgorithm in machine learning... | Continue reading
This paper demonstrates many immediate connections between adaptive controland optimization methods commonly employed in machine learning. Starting fromcommon output error formulations,... | Continue reading
With a view to bridging the gap between deep learning and symbolic AI, wepresent a novel end-to-end neural network architecture that learns to formpropositional representations with an... | Continue reading
We show that the total habitable volume in the atmospheres of cool browndwarfs with effective temperatures of $\sim 250$-$350$ K is possibly larger bytwo orders of magnitude than that of... | Continue reading
Although the popular MNIST dataset [LeCun et al., 1994] is derived from theNIST database [Grother and Hanaoka, 1995], the precise processing steps forthis derivation have been lost to time. We... | Continue reading
The vast majority of processors in the world are actually microcontrollerunits (MCUs), which find widespread use performing simple control tasks inapplications ranging from automobiles to... | Continue reading
We introduce novel variants of momentum by incorporating the variance of thestochastic loss function. The variance characterizes the confidence oruncertainty of the local features of the... | Continue reading
We demonstrate the defect-free assembly of versatile target patterns of up111 neutral atoms, building on a 361-site subset of a micro-opticalarchitecture that readily provides thousands of sites... | Continue reading
The bloom clock is a space-efficient, probabilistic data structure designedto determine the partial order of events in highly distributed systems. Thebloom clock, like the vector clock, can... | Continue reading
Recently, out-of-order execution, an important performance optimization inmodern high-end processors, has been revealed to pose a significant securitythreat, allowing information leaks across... | Continue reading
Real-time data analysis and management are increasingly critical for today`sbusinesses. SQL is the de facto lingua franca for these endeavors, yet supportfor robust streaming analysis and... | Continue reading
A core challenge for both physics and artificial intellicence (AI) issymbolic regression: finding a symbolic expression that matches data from anunknown function. Although this problem is likely... | Continue reading
We present a fully distributed collision avoidance algorithm based on convexoptimization for a team of mobile robots. This method addresses the practicalcase in which agents sense each other via... | Continue reading
Recent progress in natural language generation has raised dual-use concerns.While applications like summarization and translation are positive, theunderlying technology also might enable... | Continue reading
Convolutional Neural Networks (ConvNets) are commonly developed at a fixedresource budget, and then scaled up for better accuracy if more resources areavailable. In this paper, we systematically... | Continue reading
In society, mutual cooperation, defection, and asymmetric exploitativerelationships are common. Whereas cooperation and defection are studiedextensively in the literature on game theory,... | Continue reading