Social media monitoring by law enforcement is becoming commonplace, butlittle is known about what software packages for it do. Through public recordsrequests, we obtained log files from the... | Continue reading
Whereas it is believed that techniques such as Adam, batch normalization and,more recently, SeLU nonlinearities "solve" the exploding gradient problem, weshow that this is not the case in... | Continue reading
Social media data has become crucial to the advancement of scientificunderstanding. However, even though it has become ubiquitous, just collectinglarge-scale social media data involves a high... | Continue reading
Computer simulations are invaluable tools for scientific discovery. However,accurate simulations are often slow to execute, which limits theirapplicability to extensive parameter exploration,... | Continue reading
Semi-supervised learning (SSL) provides an effective means of leveragingunlabeled data to improve a model's performance. In this paper, we demonstratethe power of a simple combination of two... | Continue reading
Verification of modern microprocessors is a complex task that requires asubstantial allocation of resources. Despite significant progress in formalverification, the goal of complete verification... | Continue reading
Intelligent Tracking Prevention (ITP) is a privacy mechanism implemented byApple's Safari browser, released in October 2017. ITP aims to reduce thecross-site tracking of web users by limiting... | Continue reading
In this article we investigate some "unexpected" properties of the "InfinitePower Tower". \[y = f(x) = {x^{{x^{{x^{{x^{\mathinner{\mkern2mu\raise1pt\hbox{.}\mkern2mu ... | Continue reading
Automated medical image segmentation is an important step in many medicalprocedures. Recently, deep learning networks have been widely used for variousmedical image segmentation tasks, with... | Continue reading
Several variants of Convolutional Neural Networks (CNN) have been developedfor Magnetic Resonance (MR) image reconstruction. Among them, U-Net has shownto be the baseline architecture for MR... | Continue reading
Programming language design requires making many usability-related designdecisions. We explored using user-centered methods to make languages moreeffective for programmers. However, existing HCI... | Continue reading
The fruit fly Drosophila's olfactory circuit has inspired a new localitysensitive hashing (LSH) algorithm, FlyHash. In contrast with classical LSHalgorithms that produce low dimensional hash... | Continue reading
This paper is an informal survey of some of the deep connections betweenlogic and optimization. It covers George Boole's probability logic, decisiondiagrams, logic and cutting planes, first... | Continue reading
By means of a fully connected artificial neural network, we identifiedasteroids with the potential to impact Earth. The resulting instrument, namedthe Hazardous Object Identifier (HOI), was... | Continue reading
Large Transformer models routinely achieve state-of-the-art results on anumber of tasks but training these models can be prohibitively costly,especially on long sequences. We introduce two... | Continue reading
Probabilistic programming (PP) allows flexible specification of Bayesianstatistical models in code. PyMC3 is a new, open-source PP framework with anintutive and readable, yet powerful, syntax... | Continue reading
We show that the class MIP* of languages that can be decided by a classicalverifier interacting with multiple all-powerful quantum provers sharingentanglement is equal to the class RE of... | Continue reading
A recent normative turn in computer science has brought concerns aboutfairness, bias, and accountability to the core of the field. Yet recentscholarship has warned that much of this technical... | Continue reading
We present a survey of some developments in the general area ofcategory-theoretic approaches to the theory of computation, with a focus ontopics and ideas particularly close to the interests of... | Continue reading
Python currently is the dominant language in the field of Machine Learningbut is often criticized for being slow to perform certain tasks. In thisreport, we use the well-known $N$-queens puzzle... | Continue reading
Motivated by the troubling rise of political extremism and instabilitythroughout the democratic world, we present a novel mathematicalcharacterization of the nature of political representation... | Continue reading
The 'robot rights' debate, and its related question of 'robotresponsibility', invokes some of the most polarized positions in AI ethics.While some advocate for granting robots rights on a par... | Continue reading
Some changes in a recent convolution formula are performed here in order toclean it up by using more conventional notations and by making use of morereferrenced and documented components (namely... | Continue reading
A search for point-like and extended sources of cosmic neutrinos using datacollected by the ANTARES and IceCube neutrino telescopes is presented. The dataset consists of all the track-like and... | Continue reading
Women have always been underrepresented in movies and not until recently dowomen representation in movies improve. To investigate the improvement of womenrepresentation and its relationship with... | Continue reading
I show that a software framework intended primarily for training of neuralnetworks, PyTorch, is easily applied to a general function minimisation problemin science. The qualities of PyTorch of... | Continue reading
While most of the inter-galactic medium (IGM) today is permeated by ionizedhydrogen, it was largely filled with neutral hydrogen for the first 700 millionyears after the Big Bang. The process... | Continue reading
Multi-level languages and Arrows both facilitate metaprogramming, the act ofwriting a program which generates a program. The arr function required of allArrows turns arbitrary host language... | Continue reading
New consent management platforms (CMPs) have been introduced to the web toconform with the EU's General Data Protection Regulation, particularly itsrequirements for consent when companies... | Continue reading
Bloom filters (BF) are widely used for approximate membership queries over aset of elements. BF variants allow removals, sets of unbounded size or queryinga sliding window over an unbounded... | Continue reading
This paper is an attempt to explain all the matrix calculus you need in orderto understand the training of deep neural networks. We assume no math knowledgebeyond what you learned in calculus 1,... | Continue reading
Online social networks, such as Facebook and twitter, are a growingphenomenon in today's world, with various platforms providing capabilities forindividuals to collaborate through messaging and... | Continue reading
The Least-Recently Used cache replacement policy and its variants are widelydeployed in modern processors. This paper shows for the first time in detailthat the LRU states of caches can be used... | Continue reading
According to a recent survey, 2% of the U.S. population is convinced that theearth is flat. This idea is heralded by members of the Flat Earth Society, andpromulgated through Internet forums and... | Continue reading
In immersive augmented reality (IAR), users can wear a head-mounted displayto see computer-generated images superimposed to their view of the world. IARwas shown to be beneficial across several... | Continue reading
As one of the primary consumers of environmental resource, the buildingindustry faces unprecedented challenges in needing to reduce the environmentalimpact of current consumption practices. This... | Continue reading
During the last few years, intensive research efforts are being done in thefield of brain interfaces to extract neuro-information from the signalsrepresenting neuronal activities in the human... | Continue reading
I briefly summarize my personal opinions on the current status of HEP theoryand comment on the so-called "non-empiric confirmation." | Continue reading
We propose schemes for efficiently destroying privacy data in a NAND flashmemory. Generally, even if privcy data is discarded from NAND flash memories,there is a high probability that the data... | Continue reading
Cosmic-ray interactions with the solar atmosphere are expected to produceparticle showers which in turn produce neutrinos from weak decays of mesons.These solar atmospheric neutrinos (SA$ν$s)... | Continue reading
A brief overview of the models and data analyses of income, wealth,consumption distributions by the physicists, are presented here. It has beenfound empirically that the distributions of income... | Continue reading
All novae recur, but only a handful have been observed in eruption more thanonce. These systems, the recurrent novae (RNe), are among the most extremeexamples of novae. RNe have long been... | Continue reading
After years of development, FPGAs are finally making an appearance onmulti-tenant cloud servers. These heterogeneous FPGA-CPU architectures breakcommon assumptions about isolation and security... | Continue reading
In this monograph, I introduce the basic concepts of Online Learning througha modern view of Online Convex Optimization. Here, online learning refers tothe framework of regret minimization under... | Continue reading
Breast cancer remains a global challenge, causing over 1 million deathsglobally in 2018. To achieve earlier breast cancer detection, screening x-raymammography is recommended by health... | Continue reading