N
Nitish Srivastava
Researcher at Apple Inc.
Publications - 44
Citations - 47724
Nitish Srivastava is an academic researcher from Apple Inc.. The author has contributed to research in topics: Generative model & Boltzmann machine. The author has an hindex of 22, co-authored 41 publications receiving 40184 citations. Previous affiliations of Nitish Srivastava include Indian Institute of Technology Kanpur & Cornell University.
Papers
More filters
Proceedings Article
Uncertainty Weighted Actor-Critic for Offline Reinforcement Learning
Yue Wu,Shuangfei Zhai,Nitish Srivastava,Joshua M. Susskind,Jian Zhang,Ruslan Salakhutdinov,Hanlin Goh +6 more
TL;DR: In this article, uncertainty weighted actor-critic (UWAC) is proposed to detect OOD state-action pairs and down-weight their contribution in the training objectives accordingly.
Journal Article
An Attention Free Transformer
TL;DR: The Attention Free Transformers (AFT) as mentioned in this paper replaces the multi-head attention operation with the composition of element-wise multiplications/divisions and global/local pooling.
Posted Content
Unconstrained Scene Generation with Locally Conditioned Radiance Fields
Terrance DeVries,Miguel Ángel Bautista,Nitish Srivastava,Graham W. Taylor,Joshua M. Susskind +4 more
TL;DR: In this article, a generative scene network (GSN) is proposed to decompose scenes into a collection of many local radiance fields that can be rendered from a free moving camera.
Posted Content
Pointer-Chase Prefetcher for Linked Data Structures.
TL;DR: A pointer chase mechanism along with compiler hints is adopted to design a prefetcher for linked data-structures and the design is evaluated against the baseline of processor with cache in terms of performance, area and energy.
Fast Inference and Learning for Modeling Documents with a Deep Boltzmann Machine
TL;DR: This work introduces a type of Deep Boltzmann Machine that is suitable for extracting distributed semantic representations from a large unstructured collection of documents and proposes an approximate inference method that interacts with learning in a way that makes it possible to train the DBM more eciently than previously proposed methods.