arXiv subjects

Kushagra Singh

Publications and source records attributed to Kushagra Singh.

3 recordsLinked to original sources

arXiv2026

Edge Detection Framework Utilizing SOT-MTJ Bit-Cell Arrays

Traditional edge detection algorithms, foundational to computer vision, face significant challenges in energy efficiency and processing latency on conventional CMOS-based hardware. Existing algorithms, such as Canny, are computationally expensive, posing challenges in resource-constrained hardware where energy efficiency and low latency are critical. This study introduces a novel, hardware-efficient algorithm that leverages the intrinsic characteristics of magnetic tunnel junction (MTJ) devices. We present a detailed device-level analysis of an MTJ-based system for edge detection, outlining its operational cycles, including write, read, and reset methods. The algorithm's efficacy is evaluated against the standard Canny edge detection method. We provide a quantitative performance analysis, including metrics such as energy consumption and latency, which demonstrates that our proposed spintronics-based approach offers a promising solution for achieving low-power, high-speed image processing.

cond-mat.mes-hall↗

arXiv2020

How India Censors the Web

One of the primary ways in which India engages in online censorship is by ordering Internet Service Providers (ISPs) operating in its jurisdiction to block access to certain websites for its users. This paper reports the different techniques Indian ISPs are using to censor websites, and investigates whether website blocklists are consistent across ISPs. We propose a suite of tests that prove more robust than previous work in detecting DNS and HTTP based censorship. Our tests also discern the use of SNI inspection for blocking websites, which is previously undocumented in the Indian context. Using information from court orders, user reports, and public and leaked government orders, we compile the largest known list of potentially blocked websites in India. We pass this list to our tests and run them from connections of six different ISPs, which together serve more than 98% of Internet users in India. Our findings not only confirm that ISPs are using different techniques to block websites, but also demonstrate that different ISPs are not blocking the same websites.

cs.NI↗

arXiv2019

Multi-task Learning for Target-dependent Sentiment Classification

Detecting and aggregating sentiments toward people, organizations, and events expressed in unstructured social media have become critical text mining operations. Early systems detected sentiments over whole passages, whereas more recently, target-specific sentiments have been of greater interest. In this paper, we present MTTDSC, a multi-task target-dependent sentiment classification system that is informed by feature representation learnt for the related auxiliary task of passage-level sentiment classification. The auxiliary task uses a gated recurrent unit (GRU) and pools GRU states, followed by an auxiliary fully-connected layer that outputs passage-level predictions. In the main task, these GRUs contribute auxiliary per-token representations over and above word embeddings. The main task has its own, separate GRUs. The auxiliary and main GRUs send their states to a different fully connected layer, trained for the main task. Extensive experiments using two auxiliary datasets and three benchmark datasets (of which one is new, introduced by us) for the main task demonstrate that MTTDSC outperforms state-of-the-art baselines. Using word-level sensitivity analysis, we present anecdotal evidence that prior systems can make incorrect target-specific predictions because they miss sentiments expressed by words independent of target.

cs.LG↗