Nanda Harishankar Krishna

Nanda Harishankar Krishna

Auto-generated profile, based on publicly available information

Nanda Harishankar Krishna

Universite de Montreal, Mila - Quebec Artificial Intelligence Institute

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 June 2021 in npj Digital Medicine
Verified
Journal Article
Open access
PDF file
NYNaveena Yanamala
Naveena Yanamala
NKNanda H. Krishna
Nanda H. Krishna
Quincy Alexander HathawayQuincy Alexander Hathaway
Quincy Alexander Hathaway
ARAditya Radhakrishnan
Aditya Radhakrishnan
SSSrinidhi Sunkara
Srinidhi Sunkara
PSPartho P Sengupta
Partho P Sengupta
Patients with influenza and SARS-CoV2/Coronavirus disease 2019 (COVID-19) infections have a different clinical course and outcomes. We developed and validated a supervised machine learning pipeline to distinguish the two viral infections using the available vital signs and demographic dataset from the first hospital/emergency room encounters of 3883 patients who had confirmed diagnoses of influenza A/B, COVID-19 or negative laboratory test results. The models were able to achieve an area under the receiver operating characteristic curve (ROC AUC) of at least 97% using our multiclass classifier. The predictive models were externally validated on 15,697 encounters in 3125 patients available on TrinetX database that contains patient-level data from different healthcare organizations. The influenza vs COVID-19-positive model had an AUC of 98.8%, and 92.8% on the internal and external test sets, respectively. Our study illustrates the potentials of machine-learning models for accurately distinguishing the two viral infections. The code is made available at https://github.com/ynaveena/COVID-19-vs-Influenza and may have utility as a frontline diagnostic tool to aid healthcare workers in triaging patients once the two viral infections start cocirculating in the communities.

5 October 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
OCOlivier Codol
Olivier Codol
NKNanda H. Krishna
Nanda H. Krishna
GLGuillaume Lajoie
Guillaume Lajoie
MPMatthew G. Perich
Matthew G. Perich
During development, neural circuits are shaped continuously as we learn to control our bodies. The ultimate goal of this process is to produce neural dynamics that enable the rich repertoire of behaviors we perform. What begins as a series of “babbles” coalesces into skilled motor output as the brain rapidly learns to control the body. However, the nature of the teaching signal underlying this normative learning process remains elusive. Here, we test two well-established and biologically plausible theories—supervised learning (SL) and reinforcement learning (RL)—that could explain how neural circuits develop the capacity for skilled movements. We trained recurrent neural networks to control a biomechanical model of a primate arm using either SL or RL and compared the resulting neural dynamics to populations of neurons recorded from the motor cortex of monkeys performing the same movements. Intriguingly, only RL-trained networks produced neural activity that matched their biological counterparts in terms of both the geometry and dynamics of population activity. We show that this similarity with biological brains depends critically on matching biomechanical properties of the limb. Dynamical analysis on network activity revealed that our RL-trained networks operate at the “edge of chaos”, a dynamical regime known for its computational richness, greater memory capacity, and robust plasticity properties. We then demonstrated that monkeys and RL-trained networks, but not SL-trained networks, show a strikingly similar capacity for robust short-term behavioral adaptation to a movement perturbation, indicating a fundamental and general commonality in the neural control policy. Together, our results support the hypothesis that neural dynamics for behavioral control emerge through a process akin to reinforcement learning. The resulting neural circuits offer numerous advantages for adaptable behavioral control over simpler and more efficient learning rules and expand our understanding of how developmental processes shape neural dynamics.

1 December 2019
Verified
Conference Paper
Closed access
ARAbhijith Ragav
Abhijith Ragav
NKNanda Harishankar Krishna
Nanda Harishankar Krishna
NNNaveen Narayanan
Naveen Narayanan
KTKevin Thelly
Kevin Thelly
VVVineeth Vijayaraghavan
Vineeth Vijayaraghavan
Psychological stress in human beings has been on a meteoric rise over the last few years. Chronic stress can have fatal consequences such as heart disease, cancer, suicide and so on. It is thus imperative to detect stress early on to prevent health risks. In this work, we discuss efficient and accurate stress and affect detection using scalable Deep Learning methods, that can be used to monitor stress real-time on resource-constrained devices such as low-cost wearables. By making inferences on-device, we solve the issues of high latency and lack of privacy which are prevalent in cloud-based computation. Using the concept of Early Stopping - Multiple Instance Learning, we build specialized models for stress and affect detection for 3 popular datasets in the domain, that have very low inference times but high accuracy. We introduce a metric ηcomp to measure the computational savings from the use of these models. On average, our models show an absolute increase of 10% in overall accuracy over the benchmarks, computational savings of 95.39%, and an 18x reduction in inference times on a Raspberry Pi 3 Model B. This allows for efficient and accurate real-time monitoring of stress on low-cost resource-constrained devices.

1 December 2020
Verified
Conference Paper
Closed access
NKNanda H Krishna
Nanda H Krishna
PKPRAVEEN KUMAR
PRAVEEN KUMAR
RKRishi Vardhan K
Rishi Vardhan K
VVVineeth Vijayaraghavan
Vineeth Vijayaraghavan
Estimation of depth from a single image is an important scene understanding task in computer vision. With the advent of Deep Learning and Convolutional Neural Networks, staggeringly high accuracies have been achieved in this task. With advancements in model optimization, it has been possible to deploy these models on edge devices, allowing for efficient depth estimation in safety-critical applications in robots, rovers, drones and even self-driving vehicles. However, these models are susceptible to attacks from malicious adversaries, which aim to distort the output of the model for a seemingly clean image by adding minute perturbations. In the real-world scenario, the most plausible attack is the adversarial patch, which can be printed and used as a physical adversarial attack against Deep Learning models. In the case of Monocular Depth Estimation, we show that small adversarial patches, which range from 0.7% to 5% of the image size, greatly worsen model performance. It is thus essential that these models are made robust using defense mechanisms, to defend against malicious inputs while also not reducing performance on clean images. Moreover, it is essential that the defense mechanism be computationally efficient, for real-time inference on edge devices. In this work, we propose the first defense mechanism against adversarial patches for a regression network, in the context of Monocular Depth Estimation on an edge device. The defense mechanism adds very little overhead time of 38 milliseconds on a Raspberry Pi 3 Model B, maintaining performance on clean images while also achieving near clean image levels of performance on adversarial inputs.