Online tracking is a whack-a-mole game between trackers who build andmonetize behavioral user profiles through intrusive data collection, andanti-tracking mechanisms, deployed as a browser... | Continue reading
This paper does not describe a working system. Instead, it presents a singleidea about representation which allows advances made by several differentgroups to be combined into an imaginary... | Continue reading
Recently a number of empirical "universal" scaling law papers have beenpublished, most notably by OpenAI. `Scaling laws' refers to power-law decreasesof training or test error w.r.t. more data,... | Continue reading
In 1981, Mermin published a now famous paper titled, "Bringing home theatomic world: Quantum mysteries for anybody" that Feynman called, "One of themost beautiful papers in physics that I know."... | Continue reading
This paper does not describe a working system. Instead, it presents a singleidea about representation which allows advances made by several differentgroups to be combined into an imaginary... | Continue reading
The growing literature on affect among software developers mostly reports onthe linkage between happiness, software quality, and developer productivity.Understanding happiness and unhappiness in... | Continue reading
Text-to-image generation has traditionally focused on finding better modelingassumptions for training on a fixed dataset. These assumptions might involvecomplex architectures, auxiliary losses,... | Continue reading
Given a neural network, training data, and a threshold, it was known that itis NP-hard to find weights for the neural network such that the total error isbelow the threshold. We determine the... | Continue reading
Aggressive memory density scaling causes modern DRAM devices to suffer fromRowHammer, a phenomenon where rapidly activating a DRAM row can cause bit-flipsin physically-nearby rows. Recent... | Continue reading
A general framework for obtaining exact transition rate matrices forstochastic systems on networks is presented and applied to many well-knowncompartmental models of epidemiology. The state of... | Continue reading
This paper is a `spiritual child' of the 2005 lecture notes KindergartenQuantum Mechanics, which showed how a simple, pictorial extension of Diracnotation allowed several quantum features to be... | Continue reading
Similarity search finds application in specialized database systems handlingcomplex data such as images or videos, which are typically represented byhigh-dimensional features and require... | Continue reading
We explore a new class of end-to-end learnable models wherein data processingnodes (or network layers) are defined in terms of desired behavior rather thanan explicit forward function.... | Continue reading
We present lambda layers -- an alternative framework to self-attention -- forcapturing long-range interactions between an input and structured contextualinformation (e.g. a pixel surrounded by... | Continue reading
We present the Open Predicate Query Language (OPQL); a method forconstructing a virtual KB (VKB) trained entirely from text. Large KnowledgeBases (KBs) are indispensable for a wide-range of... | Continue reading
The proliferation of peer-to-peer (P2P) file sharing protocols is due totheir efficient and scalable methods for data dissemination to numerous users.But many of these networks have no... | Continue reading
The task of programming a quantum computer is just as strange as quantummechanics itself. But it now looks like a simple 3D puzzle may be the futuretool of quantum software engineers. | Continue reading
Transformers have emerged as a powerful tool for a broad range of naturallanguage processing tasks. A key component that drives the impressiveperformance of Transformers is the self-attention... | Continue reading
The allocation of venture capital is one of the primary factors determiningwho takes products to market, which startups succeed or fail, and as such whogets to participate in the shaping of our... | Continue reading
We present a novel system that gets as an input video frames of a musicianplaying the piano and generates the music for that video. Generation of musicfrom visual cues is a challenging problem... | Continue reading
The majority of smartphones either run iOS or Android operating systems. Thishas created two distinct ecosystems largely controlled by Apple and Google -they dictate which applications can run,... | Continue reading
The ability to use symbols is the pinnacle of human intelligence, but has yetto be fully replicated in machines. Here we argue that the path towardssymbolically fluent artificial intelligence... | Continue reading
Inference attacks against Machine Learning (ML) models allow adversaries tolearn information about training data, model parameters, etc. While researchershave studied these attacks thoroughly,... | Continue reading
We present TEyeD, the world's largest unified public data set of eye imagestaken with head-mounted devices. TEyeD was acquired with seven differenthead-mounted eye trackers. Among them, two eye... | Continue reading
The capabilities of natural language models trained on large-scale data haveincreased immensely over the past few years. Downstream applications are atrisk of inheriting biases contained in... | Continue reading
We analyse a megasatellite settlement built from Ceres materials in highCeres orbit. Ceres is selected because it has nitrogen, which is necessary foran earthlike atmosphere. To have $1 g$... | Continue reading
We describe a formal proof of the independence of the continuum hypothesis($\mathsf{CH}$) in the Lean theorem prover. We use Boolean-valued models togive forcing arguments for both directions,... | Continue reading
Standard training techniques for neural networks involve multiple sources ofrandomness, e.g., initialization, mini-batch ordering and in some cases dataaugmentation. Given that neural networks... | Continue reading
Arb is a C library for arbitrary-precision interval arithmetic using themidpoint-radius representation, also known as ball arithmetic. It supports realand complex numbers, polynomials, power... | Continue reading
The intertwined processes of learning and evolution in complex environmentalniches have resulted in a remarkable diversity of morphological forms.Moreover, many aspects of animal intelligence... | Continue reading
In this paper, we propose a clearing model for prices in a financial marketsdue to margin calls on short sold assets. In doing so, we construct an explicitformulation for the prices that would... | Continue reading
According to a hard-steps model of advanced life timing, humans seempuzzlingly early. We offer an explanation: an early deadline is set by 'grabby'civilizations (GC), who expand rapidly, never... | Continue reading
Large-scale model training has been a playing ground for a limited fewrequiring complex model refactoring and access to prohibitively expensive GPUclusters. ZeRO-Offload changes the large model... | Continue reading
We study empirical scaling laws for transfer learning between distributionsin an unsupervised, fine-tuning setting. When we train increasingly largeneural networks from-scratch on a fixed-size... | Continue reading
We study a model of user decision-making in the context of recommendersystems via numerical simulation. Our model provides an explanation for thefindings of Nguyen, et. al (2014), where, in... | Continue reading
When working with distant collaborators on the same documents, one often usesa version control system, which is a program tracking the history of files andhelping importing modifications brought... | Continue reading
Artificial Intelligence (AI) is being hailed as the latest example of aGeneral Purpose Technology that could transform productivity and help tackleimportant societal challenges. This outcome is... | Continue reading
If advanced civilizations appear in the universe with an ability and desireto expand, the entire universe can become saturated with life on a shorttimescale, even if such expanders appear... | Continue reading
Training machine learning models requires feeding input data for models toingest. Input pipelines for machine learning jobs are often challenging toimplement efficiently as they require reading... | Continue reading
The mushroom body of the fruit fly brain is one of the best studied systemsin neuroscience. At its core it consists of a population of Kenyon cells, whichreceive inputs from multiple sensory... | Continue reading
There is a rapidly growing literature on dark patterns, user interfacedesigns -- typically related to shopping or privacy -- that researchers deemproblematic. Recent work has been predominantly... | Continue reading
Cryptographic protocols have been widely used to protect the user's privacyand avoid exposing private information. QUIC (Quick UDP Internet Connections),as an alternative to traditional HTTP,... | Continue reading
One of the major optimizations employed in deep learning frameworks is graphrewriting. Production frameworks rely on heuristics to decide if rewrite rulesshould be applied and in which order.... | Continue reading
A large-scale, up-to-date analysis of Computer Science literature (11.8Mpapers through 2019) reveals that, if trends from the last 50 years continue,parity between the number of male and female... | Continue reading
This book is an introduction to program analysis that is meant to beconsiderably more elementary than our advanced book Principles of ProgramAnalysis (Springer, 2005). Rather than using flow... | Continue reading
The SIMT execution model is commonly used for general GPU development. CUDAand OpenCL developers write scalar code that is implicitly parallelized bycompiler and hardware. On Intel GPUs,... | Continue reading
Cloud gaming is a new class of services that promises to revolutionize thevideogame market. It allows the user to play a videogame with basic equipmentwhile using a remote server for the actual... | Continue reading