Yeshwanth Nagaraj

Yeshwanth Nagaraj

Karnataka, India
4K followers 500+ connections

Join to follow

Articles by Yeshwanth

See all articles

Activity

4K followers

See all activities

Experience

  • Microsoft Graphic

    Microsoft

  • -

    Bengaluru, Karnataka, India

  • -

    Bengaluru, Karnataka, India

  • -

    Bengaluru, Karnataka

  • -

    Bengaluru Area, India

  • -

    Bengaluru Area, India

  • -

    Bangaon Area, India

Education

Licenses & Certifications

Join now to see all certifications

Volunteer Experience

  • MSPL Limited Graphic

    Volunteer

    MSPL Limited

    - Present 20 years 6 months

    Environment

    Mass Afforestation Program on World Evironment Day

Publications

  • Gradient Guard: Robust Federated Learning using Saliency Maps

    IEEE

    This paper presents a novel approach to counter Directed Deviation Attacks (DDA) in the domain of Federated Learning (FL). DDAs exploit gradient manipulation, disrupting model learning and increasing test error rates. This study introduces a novel defense mechanism employing Saliency Maps, a tool highlighting influential input regions to detect gradient anomalies caused by malicious clients.Existing defenses struggle against DDAs, prompting exploration of an alternative solution. By…

    This paper presents a novel approach to counter Directed Deviation Attacks (DDA) in the domain of Federated Learning (FL). DDAs exploit gradient manipulation, disrupting model learning and increasing test error rates. This study introduces a novel defense mechanism employing Saliency Maps, a tool highlighting influential input regions to detect gradient anomalies caused by malicious clients.Existing defenses struggle against DDAs, prompting exploration of an alternative solution. By quantitatively measuring Structural Similarity Index (SSI) between Saliency Maps of benign and potentially malicious client updates, abnormal gradient patterns can be swiftly identified. This method safeguards FL models by isolating contributions of suspicious clients. Its efficacy across diverse algorithms, architectures, and datasets is demonstrated.Empirical evaluations reveal superiority, with the proposed approach ensuring Byzantine-robustness against up to 50% malicious clients, compared to traditional defenses 20-30% limit and recent work, FLAIR which offers byzantine-robustness upto a malicious client percentage of 45%. This paper introduces a pioneering technique combining Saliency Maps and SSI to protect FL models and underscores the need for proactive measures in the evolving landscape of decentralized learning.The rise of Federated Learning (FL) brings decentralized learning to the forefront, enabling efficient and private model training across devices. However, FL’s decentralized nature exposes it to adversarial challenges, including Directed Deviation Attacks (DDA). DDAs manipulate gradients to divert models from optimal paths, increasing test errors. Traditional defenses fall short against DDAs, necessitating innovative solutions.This research harnesses Saliency Maps, traditionally used to understand model behavior, for defense against DDAs. Coupled with the Structural Similarity Index (SSI), it provides a dual visual and quantitative strategy...

    Other authors
    See publication

  • GradClassify: Securing Federated Learning Using Open Set Classification on Gradients

    IEEE

    The proposed method introduces a novel approach to enhancing security in Federated Learning (FL) systems by mitigating model poisoning attacks. The core innovation lies in the incorporation of an Open Set Classifier (OSC) that scrutinizes gradient updates during the learning process. This classifier is designed to recognize benign gradient updates while flagging any deviations from the norm as potentially malicious. By doing so, the system aims to maintain the integrity of the federated model…

    The proposed method introduces a novel approach to enhancing security in Federated Learning (FL) systems by mitigating model poisoning attacks. The core innovation lies in the incorporation of an Open Set Classifier (OSC) that scrutinizes gradient updates during the learning process. This classifier is designed to recognize benign gradient updates while flagging any deviations from the norm as potentially malicious. By doing so, the system aims to maintain the integrity of the federated model without compromising on the learning efficiency. The architecture of the OSC for gradient updates is elaborately discussed, making it a comprehensive solution for securing FL environments against various types of attacks.

    Other authors
    See publication

  • AP-TRL: Augmenting Real-Time Personalization with Transformer Reinforcement Learning

    IEEE

    In the digital era, understanding user behavior in real-time and providing immediate personalization can significantly enhance the user experience and engagement. This paper introduces a novel approach that integrates the transformer architecture with reinforcement learning (RL) for real-time user behavior tracking, and recommendation. This paper demonstrates how this hybrid model can efficiently categorize user actions, predict future behaviors, and personalize content in real-time…

    In the digital era, understanding user behavior in real-time and providing immediate personalization can significantly enhance the user experience and engagement. This paper introduces a novel approach that integrates the transformer architecture with reinforcement learning (RL) for real-time user behavior tracking, and recommendation. This paper demonstrates how this hybrid model can efficiently categorize user actions, predict future behaviors, and personalize content in real-time. Experimental results show that our model outperforms traditional methods, with a marked improvement in accuracy and response time.

    Other authors
    See publication

  • AST-MHSA : Code Summarization using Multi-Head Self-Attention

    Cornell University

    Code summarization aims to generate concise natural language descriptions for source code. The prevailing approaches adopt transformer-based encoder-decoder architectures, where the Abstract Syntax Tree (AST) of the source code is utilized for encoding structural information. However, ASTs are much longer than the corresponding source code, and existing methods ignore this size constraint by directly feeding the entire linearized AST into the encoders. This simplistic approach makes it…

    Code summarization aims to generate concise natural language descriptions for source code. The prevailing approaches adopt transformer-based encoder-decoder architectures, where the Abstract Syntax Tree (AST) of the source code is utilized for encoding structural information. However, ASTs are much longer than the corresponding source code, and existing methods ignore this size constraint by directly feeding the entire linearized AST into the encoders. This simplistic approach makes it challenging to extract truly valuable dependency relations from the overlong input sequence and leads to significant computational overhead due to self-attention applied to all nodes in the AST.
    To address this issue effectively and efficiently, we present a model, AST-MHSA that uses multi-head attention to extract the important semantic information from the AST. The model consists of two main components: an encoder and a decoder. The encoder takes as input the abstract syntax tree (AST) of the code and generates a sequence of hidden states. The decoder then takes these hidden states as input and generates a natural language summary of the code.
    The multi-head attention mechanism allows the model to learn different representations of the input code, which can be combined to generate a more comprehensive summary. The model is trained on a dataset of code and summaries, and the parameters of the model are optimized to minimize the loss between the generated summaries and the ground-truth summaries.

    Other authors
    See publication

  • SPARSE N-TRAIN SET GAN PIPELINE TO BUILD ARTIFACT QUANTIFIER

    ip.com

    The present disclosure proposes a novel approach i.e., a classical GAN pipeline approach to generate data of larger dataset and then train the image segmentation model by proper data to evaluate the generated data. The proposed method mainly involves a data augmentation process to generate data from small datasets, a data filtering process to get only selective portion of the generated data, and an image segmentation process to obtain a trained model to thereby segment a portion of the scanned…

    The present disclosure proposes a novel approach i.e., a classical GAN pipeline approach to generate data of larger dataset and then train the image segmentation model by proper data to evaluate the generated data. The proposed method mainly involves a data augmentation process to generate data from small datasets, a data filtering process to get only selective portion of the generated data, and an image segmentation process to obtain a trained model to thereby segment a portion of the scanned image indicating a particular artifact in it and also percentage of the whole image thus quantifying the image quality.

    See publication

Courses

  • Advanced Communication

    78.4

  • Advanced Microprocessor

    78.4

  • Computer Communication Networks

    74.4

  • DSP Algorithms and Architecture

    75.4

  • Embedded System Design

    74.4

  • Industrial Automation

    76.8%

  • Optical Fiber Communication

    57.6

  • Power Electronics

    58

  • Real Time Systems

    72

  • VHDL Programming

    85.6

Projects

  • CricketSimulator

    -

    https://github.com/yeshsurya/CricketSimulator
    Has a simplistic weighted random generator. A good proto for weighted random number generator. Commentary will be generated for each ball. Any team can win ! Keep running multiple times to get different results. [ You got to run this! ]

  • MyPythonScripts

    -

    https://github.com/yeshsurya/MyPythonScripts
    Desktop Cleaner

    Running Script :
    1.Install python here : https://www.python.org/downloads/
    2.Open command prompt(windows) or terminal (Mac)
    3.Navigate to location whre desktopCleaner.py is present
    4.Type "python desktopCleaner.py" ( Assuming python got installed successfully ])

    Run this script to move all files to a Folder named "DesktopStack".
    Folders will be created inside "DesktopStack" based on uniqe file…

    https://github.com/yeshsurya/MyPythonScripts
    Desktop Cleaner

    Running Script :
    1.Install python here : https://www.python.org/downloads/
    2.Open command prompt(windows) or terminal (Mac)
    3.Navigate to location whre desktopCleaner.py is present
    4.Type "python desktopCleaner.py" ( Assuming python got installed successfully ])

    Run this script to move all files to a Folder named "DesktopStack".
    Folders will be created inside "DesktopStack" based on uniqe file extensions found on the desktop.
    All the files will be moved based on extension name into those respecitve folders.
    Slide Share Slides Downloader
    Utility to download all the slides from the slideshare page.
    Directory Maker
    Given a text file with list of Names delimited by carriage return, utility will create a folder for each element in the list.
    Remote SSH Command Executer
    Project vision is to pick a file containing list of hosts and command to be executed and print the required result to output file
    Make Slides From Video
    Get only snaps from video as and when only there is change in them. Helps get notes of a online course generated automatically.
    Latent Semantic Analysis
    In this example, we use the LSA (Latent Semantic Analysis) summarization algorithm from sumy. You can adjust the num_sentences parameter to specify the desired length of the summary.
    GenCaptionsForVideo
    To generate captions for a video input using Python, you can utilize the Deep Learning-based video captioning model and the moviepy library for video processing.

Honors & Awards

  • 1st Rank Holder

    KLM Awards

    For First in SSLC

  • Drawing Grade 1 - KSEEB

    KSEEB

    KSEEB issues a certicate after conducting various drawing tests to cerity medium to advanced drawing skill expertise.

  • GEM( Go The Extra Mile) Award

    Aptean

    For outstanding performance from day 1

Test Scores

  • DCET,Entrance Test From Diploma to Engineering

    Score: 5

    State 5th Rank in Entrance Test in Electronics and Communication. 22nd Rank in All Engineering Streams List.

Organizations

  • OpenStack

    Member

    - Present

    Part of Open Stack community, pushing forwared OpenStack code base.

Recommendations received

5 people have recommended Yeshwanth

Join now to view

View Yeshwanth’s full profile

  • See who you know in common
  • Get introduced
  • Contact Yeshwanth directly
Join to view full profile

Other similar profiles

Explore more posts

Explore collaborative articles

We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.

Explore More

Add new skills with these courses

See all courses