Given a node-attributed graph, and a graph task (link prediction or node classification), can we tell if a graph neural network (GNN) will perform well? More specifically, do the graph structure and the node features carry enough usable information for the task? Our goals are (1) to develop a fast…
Amazon IIT–Bombay AI-ML Initiative seeks to advance artificial intelligence and machine learning research within speech, language, and multimodal-AI domains.
An effective approach to design automated Question Answering (QA) systems is to efficiently retrieve answers from pre-computed databases containing question/answer pairs. One of the main challenges to this design is the lack of training/testing data. Existing resources are limited in size and…
Road attributes play a pivotal role in digital maps, providing critical information for various routing and planning applications that aim to create a safe and efficient traffic environment. While some road attributes are available in existing map data such as OpenStreetMap [3], these sources may…
Deep learning training compilers accelerate and achieve more resource-efficient training. We present a deep learning compiler for training consisting of three main features, a syncfree optimizer, compiler caching and multi-threaded execution. We demonstrate speedups for common language and vision…
Neural network implementations have predominantly been a black box lacking both in interpretability and estimation of uncertainty. In this study, we propose a novel causal attribution methodology for mixture density networks wherein we outline a framework to compute the causal effect of each…
We consider the problem of online service with delay on a general metric space, first presented by Azar, Ganesh, Ge and Panigrahi (STOC 2017). The best known randomized algorithm for this prob-lem, by Azar and Touitou (FOCS 2019), is ???? (log2 ????)-competitive, where ???? is the number of points in the…
We introduce a new distribution-free multi-horizon forecast. As such, it integrates product-level forecast at fixed lead times, spans, and quantiles with the ability for a user to request a forecast at any other lead time, span or quantile via piecewise linear interpolation with exponential…
A complex logic query in a knowledge graph refers to a query expressed in logic form that conveys a complex meaning, such as where did the Canadian Turing award winner graduate from? Knowledge graph reasoning-based applications, such as dialogue systems and interactive search engines, rely on the…
We advance the study of incentivized bandit exploration, in which arm choices are viewed as recommendations and are required to be Bayesian incentive compatible. Recent work of (Sellke & Slivkins, 2022) has shown that for the special case of independent arms, after collect-ing enough initial…
Commercial search engines use different semantic models to augment lexical matches. These models provide candidate items for a user’s query from a target space of millions to billions of items. Models with different inductive biases provide relatively different predictions, making it desirable to…
Hyperparameter optimization (HPO) and neural architecture search (NAS) are methods of choice to obtain the best-in-class machine learning models, but in practice they can be costly to run. When models are trained on large datasets, tuning them with HPO or NAS rapidly becomes prohibitively expensive…
We introduce a variant of the endpoint (EP) detection problem in automatic speech recognition (ASR), which we call the end-of-speech (EOS) estimation. Given an utterance, EOS estimation aims to identify the timestamp when the utterance waveform has fully decayed and is then used to measure the EP…
Methods for controlling the outputs of large generative models and integrating symbolic reasoning with machine learning are among the conference’s hot topics.
Prem Natarajan, Alexa AI vice president, and Michael Kearns, an Amazon Scholar, discuss fairness, accountability, transparency, and ethics topics applied to machine learning, automation, robotics, and space themes.
Automating updates to machine learning systems is an important but understudied challenge in AutoML. The high model variance of many cutting-edge deep learning architectures means that retraining a model provides no guarantee of accurate inference on all sample types. To address this concern, we…
We provide a differentially private algorithm for producing synthetic data simultaneously useful for multiple tasks: marginal queries and multitask machine learning (ML). A key innovation in our algorithm is the ability to directly handle numerical features, in contrast to a number of related prior…
Adaptive experimental design methods are increasingly being used in industry as a tool to boost testing throughput or reduce experimentation cost relative to traditional A/B/N testing methods. This paper shares lessons learned regarding the challenges and pitfalls of naively using adaptive…