Source author record

S. Narayanan

S. Narayanan appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

quant-ph Artificial Intelligence astro-ph.CO astro-ph.GA astro-ph.IM Computation and Language Distributed, Parallel, and Cluster Computing Machine Learning Networking and Internet Architecture

Catalog footprint

What is connected

5works

9topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2026arXiv

Resilient AI Supercomputer Networking using MRC and SRv6

Tail latency dominates the performance of synchronous pretraining jobs when running at very large scales. We describe a three-pronged approach: (1) a new RDMA-based transport protocol, MRC, sprays across many paths and actively load-balances between them, eliminating the issue of flow collisions (2) the use of multi-plane Clos topologies to get the benefits of high switch radix and redundancy, allowing training clusters well over 100K GPUs to be built as two-tier topologies while increasing physical redundancy, and (3) the use of static source-routing using SRv6 to allow MRC the freedom to bypass failures by itself. We describe our experiences running MRC and static SRv6 routing in production in OpenAI and Microsoft's largest training clusters, where it has been used to train the latest frontier models. We demonstrate how MRC allows AI training jobs to ride out many network failures that previously would have interrupted training.

preprint2020arXiv

Affective Conditioning on Hierarchical Networks applied to Depression Detection from Transcribed Clinical Interviews

In this work we propose a machine learning model for depression detection from transcribed clinical interviews. Depression is a mental disorder that impacts not only the subject's mood but also the use of language. To this end we use a Hierarchical Attention Network to classify interviews of depressed subjects. We augment the attention layer of our model with a conditioning mechanism on linguistic features, extracted from affective lexica. Our analysis shows that individuals diagnosed with depression use affective language to a greater extent than not-depressed. Our experiments show that external affective information improves the performance of the proposed architecture in the General Psychotherapy Corpus and the DAIC-WoZ 2017 depression datasets, achieving state-of-the-art 71.6 and 68.6 F1 scores respectively.

preprint2016arXiv

Brute-Force Mapmaking with Compact Interferometers: A MITEoR Northern Sky Map from 128 MHz to 175 MHz

We present a new method for interferometric imaging that is ideal for the large fields of view and compact arrays common in 21 cm cosmology. We first demonstrate the method with simulations for two very different low frequency interferometers, the Murchison Widefield Array (MWA) and the MIT Epoch of Reionization (MITEoR) Experiment. We then apply the method to the MITEoR data set collected in July 2013 to obtain the first northern sky map from 128 MHz to 175 MHz at about 2 degree resolution, and find an overall spectral index of -2.73+/-0.11. The success of this imaging method bodes well for upcoming compact redundant low-frequency arrays such as HERA. Both the MITEoR interferometric data and the 150 MHz sky map are publicly available at http://space.mit.edu/home/tegmark/omniscope.html.

preprint2010arXiv

Fabrication and heating rate study of microscopic surface electrode ion traps

We report heating rate measurements in a microfabricated gold-on-sapphire surface electrode ion trap with trapping height of approximately 240 micron. Using the Doppler recooling method, we characterize the trap heating rates over an extended region of the trap. The noise spectral density of the trap falls in the range of noise spectra reported in ion traps at room temperature. We find that during the first months of operation the heating rates increase by approximately one order of magnitude. The increase in heating rates is largest in the ion loading region of the trap, providing a strong hint that surface contamination plays a major role for excessive heating rates. We discuss data found in the literature and possible relation of anomalous heating to sources of noise and dissipation in other systems, namely impurity atoms adsorbed on metal surfaces and amorphous dielectrics.

preprint2009arXiv

Wiring up trapped ions to study aspects of quantum information

There has been much interest in developing methods for transferring quantum information. We discuss a way to transfer quantum information between two trapped ions through a wire. The motion of a trapped ion induces oscillating charges in the trap electrodes. By sending this current to the electrodes of a nearby second trap, the motions of ions in the two traps are coupled. We investigate the electrostatics of a set-up where two separately trapped ions are coupled through an electrically floating wire. We also discuss experimental issues, including possible sources of decoherence.