Showing 1–2 of 2 results for author: Singh, A A

TransEdge: Supporting Efficient Read Queries Across Untrusted Edge Nodes

Authors: Abhishek A. Singh, Aasim Khan, Sharad Mehrotra, Faisal Nawab

Abstract: We propose Transactional Edge (TransEdge), a distributed transaction processing system for untrusted environments such as edge computing systems. What distinguishes TransEdge is its focus on efficient support for read-only transactions. TransEdge allows reading from different partitions consistently using one round in most cases and no more than two rounds in the worst case. TransEdge design is ce… ▽ More We propose Transactional Edge (TransEdge), a distributed transaction processing system for untrusted environments such as edge computing systems. What distinguishes TransEdge is its focus on efficient support for read-only transactions. TransEdge allows reading from different partitions consistently using one round in most cases and no more than two rounds in the worst case. TransEdge design is centered around this dependency tracking scheme including the consensus and transaction processing protocols. Our performance evaluation shows that TransEdge's snapshot read-only transactions achieve an 9-24x speedup compared to current byzantine systems. △ Less

Submitted 15 February, 2023; originally announced February 2023.

mask-Net: Learning Context Aware Invariant Features using Adversarial Forgetting (Student Abstract)

Authors: Hemant Yadav, Atul Anshuman Singh, Rachit Mittal, Sunayana Sitaram, Yi Yu, Rajiv Ratn Shah

Abstract: Training a robust system, e.g.,Speech to Text (STT), requires large datasets. Variability present in the dataset such as unwanted nuisances and biases are the reason for the need of large datasets to learn general representations. In this work, we propose a novel approach to induce invariance using adversarial forgetting (AF). Our initial experiments on learning invariant features such as accent o… ▽ More Training a robust system, e.g.,Speech to Text (STT), requires large datasets. Variability present in the dataset such as unwanted nuisances and biases are the reason for the need of large datasets to learn general representations. In this work, we propose a novel approach to induce invariance using adversarial forgetting (AF). Our initial experiments on learning invariant features such as accent on the STT task achieve better generalizations in terms of word error rate (WER) compared to the traditional models. We observe an absolute improvement of 2.2% and 1.3% on out-of-distribution and in-distribution test sets, respectively. △ Less

Submitted 18 October, 2021; v1 submitted 25 November, 2020; originally announced November 2020.