We introduce the idea that using optimal classification trees (OCTs) andoptimal classification trees with-hyperplanes (OCT-Hs), interpretable machinelearning algorithms developed in [BD17,... | Continue reading
Contemporary domain adaptation methods are very effective at aligning featuredistributions of source and target domains without any target supervision.However, we show that these techniques... | Continue reading
It is well-known that the expressivity of a neural network depends on itsarchitecture, with deeper networks expressing more complex functions. In thecase of networks that compute piecewise... | Continue reading
Deep learning for medical image classification faces three major challenges:1) the number of annotated medical images for training are usually small; 2)regions of interest (ROIs) are relatively... | Continue reading
In this work, we address the task of natural image generation guided by aconditioning input. We introduce a new architecture called conditionalinvertible neural network (cINN). The cINN combines... | Continue reading
By borrowing the wisdom of human in gaze following, we propose a two-stagesolution for gaze point prediction of the target persons in a scene.Specifically, in the first stage, both head image... | Continue reading
Recent advances in deep learning have pushed the performances of visualsaliency models way further than it has ever been. Numerous models in theliterature present new ways to design neural... | Continue reading
We present LumièreNet, a simple, modular, and completely deep-learningbased architecture that synthesizes, high quality, full-pose headshot lecturevideos from instructor's new audio narration... | Continue reading
Climate change is one of the greatest challenges facing humanity, and we, asmachine learning experts, may wonder how we can help. Here we describe howmachine learning can be a powerful tool in... | Continue reading
Image deblurring is a fundamental and challenging low-level vision problem.Previous vision research indicates that edge structure in natural scenes is oneof the most important factors to... | Continue reading
In this age of social media, people often look at what others are wearing. Inparticular, Instagram and Twitter influencers often provide images ofthemselves wearing different outfits and their... | Continue reading
Text detection plays a critical role in the whole procedure of textualinformation extraction and understanding. On a high note, recent years haveseen a surge in the high recall text detectors in... | Continue reading
Software complexity has increased over the years. One common way to tacklethis complexity during development is to encapsulate features into a sharedlibrary. This allows developers to reuse... | Continue reading
Crowd counting aims to count the number of instantaneous people in a crowdedspace, which plays an increasingly important role in the field of publicsafety. More and more researchers have already... | Continue reading
We introduce a computationally-efficient CNN micro-architecture Slim Moduleto design a lightweight deep neural network Slim-Net for face attributeprediction. Slim Modules are constructed by... | Continue reading
Deep learning frameworks commonly implement convolution operators withGEMM-based algorithms. In these algorithms, convolution is implemented on topof matrix-matrix multiplication (GEMM)... | Continue reading
Since its conception in 2006, differential privacy has emerged as thede-facto standard in data privacy, owing to its robust mathematical guarantees,generalised applicability and rich body of... | Continue reading
Products and coproducts may be recognized as morphisms in a monoidal tensorcategory of vector spaces. To gain invariant data of these morphisms, we canuse singular value decomposition which... | Continue reading
A decentralized key management infrastructure (DKMI) that uses the designprinciple of minimally sufficient means is presented. The primary keymanagement operation is key rotation via a novel key... | Continue reading
This paper presents a case study on applying two model checkers, SPIN andDIVINE, to verify key properties of a C++ software framework, known as ADAPRO,originally developed at CERN. SPIN was used... | Continue reading
Open access WiFi hotspots are widely deployed in many public places,including restaurants, parks, coffee shops, shopping malls, trains, airports,hotels, and libraries. While these hotspots... | Continue reading
We address the task of assessing discourse coherence, an aspect of textquality that is essential for many NLP tasks, such as summarization andlanguage assessment. We propose a hierarchical... | Continue reading
Linguistic similarity is multi-faceted. For instance, two words may besimilar with respect to semantics, syntax, or morphology inter alia. Continuousword-embeddings have been shown to capture... | Continue reading
Pinterest is a popular Web application that has over 250 million activeusers. It is a visual discovery engine for finding ideas for recipes, fashion,weddings, home decoration, and much more. In... | Continue reading
Several recent works have shown how highly realistic human head images can beobtained by training convolutional neural networks to generate them. In orderto create a personalized talking head... | Continue reading
This paper investigates the application of machine learning (ML) techniquesto enable intelligent systems to learn multi-party turn-taking models fromdialogue logs. The specific ML task consists... | Continue reading
In recent years, the biggest advances in major Computer Vision tasks, such asobject recognition, handwritten-digit identification, facial recognition, andmany others., have all come through the... | Continue reading
Programmers of cryptographic applications written in C need to avoid commonmistakes such as sending private data over public channels, modifying trusteddata with untrusted functions, or... | Continue reading
Recent studies have shown that aggregate CPU usage and power consumptiontraces on smartphones can leak information about applications running on thesystem or websites visited. In response,... | Continue reading
We extend two kinds of causal models, structural equation models andsimulation models, to infinite variable spaces. This enables a semantics forconditionals founded on a calculus of... | Continue reading
A central capability of intelligent systems is the ability to continuouslybuild upon previous experiences to speed up and enhance learning of new tasks.Two distinct research paradigms have... | Continue reading
Many machine learning tasks require sampling a subset of items from acollection based on a parameterized distribution. The Gumbel-softmax trick canbe used to sample a single item, and allows for... | Continue reading
Model-based reinforcement learning (MBRL) is widely seen as having thepotential to be significantly more sample efficient than model-free RL.However, research in model-based RL has not been very... | Continue reading
Perspective taking is the ability to take the point of view of another agent.This skill is not unique to humans as it is also displayed by other animalslike chimpanzees. It is an essential... | Continue reading
Interval type-2 (IT2) fuzzy systems have become increasingly popular in thelast 20 years. They have demonstrated superior performance in manyapplications. However, the operation of an IT2 fuzzy... | Continue reading
Values are things that are important to us. Actions activate values - theyeither go against our values or they promote our values. Values themselves caneither be conforming or conflicting... | Continue reading
With only five photographs of the Sun at different dates we show that themass of Sun can be calculated by using a telescope, a camera, and the thirdKepler's law. With the photographs we are able... | Continue reading
Rule-based systems play a critical role in health and safety, where policiescreated by experts are usually formalised as rules. When dealing withincreasingly large and dynamic sources of data,... | Continue reading
Cycloids, hipocycloids and epicycloids have an often forgotten commonproperty: they are homothetic to their evolutes. But what if use convexsymmetric polygons as unit balls, can we define... | Continue reading
Balancing an ever growing strategic game of high complexity, such asHearthstone is a complex task. The target of making strategies diverse andcustomizable results in a delicate intricate system.... | Continue reading
We propose and investigate feasibility of a novel task that consists infinding e-sports talent using multimodal Twitch chat and video stream data. Inthat, we focus on predicting the ranks of... | Continue reading
Conventionally, model-based reinforcement learning (MBRL) aims to learn aglobal model for the dynamics of the environment. A good model can potentiallyenable planning algorithms to generate a... | Continue reading
Recent work has considered the problem of extending to the case of iteratedbelief change the so-called `Harper Identity' (HI), which defines single-shotcontraction in terms of single-shot... | Continue reading
Playlists have become a significant part of our listening experience becauseof the digital cloud-based services such as Spotify, Pandora, Apple Music.Owing to the meteoric rise in the usage of... | Continue reading
In the recent years, there is a growing interest in semi-supervised learning,since, in many learning tasks, there is a plentiful supply of unlabeled data,but insufficient labeled ones. Hence,... | Continue reading
There has been an increasing interest in semi-supervised learning in therecent years because of the great number of datasets with a large number ofunlabeled data but only a few labeled samples.... | Continue reading
These are lecture notes from a course offered at the Bangalore School onStatistical Physics - X, during 17-28 June 2019, [https://www.icts.res.in/program/bssp2019 ] at International centre... | Continue reading