Ishika Singh

Ishika Singh

Undergrad, Computer Science majors

IIT Kanpur

Biography

I am a final year undergrad at IIT Kanpur. I am fortunate to have worked with Prof. Ashutosh Modi (IIT-K), Prof. Nisheeth Srivastava (IIT-K), and Prof. Pengtao Xie ( UC San Diego). I have spent one of my summers interning at Adobe Research. I have mostly worked in NLP and theoretical ML (see publications). My research interests are across NLP, RL, and ML theory; primarily around why (how) language is necessary for (can be utilized in) robot learning. I think that problems under Embodied AI are yet more foundational than applied, and hold deep-rooted benefits for the society.

I am also a part of Brain & Cognitive Society (BCS) at IIT-K, currently leading the group, where I organize interesting talks and collaborate/mentor projects. I have also been an Academic Mentor to freshers for 1st year Mathematics Courses offered at IIT-K.

In my leisure time, I participated in a competition ( TechnoUtsav 2.0) organized by Deloitte, and my idea on “Intelligent Energy Harvesting” won the Grand Prize. My team was invited to US India Analytics Summit 2019, where I also met my youngest inspiration Tanmay Bakshi. In 2020, Adobe Research selected me for Adobe India WIT Scholarship. I have spent sometime in electrochemical research as well under Prof. Raj Pala. He has been a constant source of encouragement and knowledge (and fresh/smart jokes :P) to me.

I also spend time writing about my philosophical thoughts and have a predilection towards painting (traditional art). To know more, contact me on your favorite interface from the menu on your left. I am most efficient with emails and twitter.

Interests

  • RoboNLP, Embodied AI
  • Distributed Learning and Optimization
  • Human-Centred AI

Education

  • B.Tech., double major in Computer Science and Chemical Engineering, 2016 - 2021

    Indian Institute of Technology Kanpur, India

Updates

Sept, 2020 | Work on Affective Text Generation accepted as long paper at COLING’20. [ arXiv]
Sept, 2020 | Adobe released a blog on my internship at Adobe Research, India! [ Read here]
Jun, 2020 | Work on Differentially-private Federated Neural Architecture Search accepted at FL-ICML’20. [ arXiv]
Dec, 2019 | Received Adobe India WIT Scholarship 2020. [ Adobe’s Blog][ Getting-through-it Advice!]

Publications

Adapting a Language Model for Controlled Affective Text Generation

Ishika Singh*, Ahsan Barkati*, Tushar Goswamy, Ashutosh Modi
International Conference on Computational Linguistics (COLING) 2020
Appeared at Wordplay: When Language Meets Games @ NeurIPS 2020
arXiv / talk / code

Differentially-private Federated Neural Architecture Search

Ishika Singh*, Haoyi Zhou*, Kunlin Yang, Meng Ding, Bill Lin, Pengtao Xie
International Workshop on Federated Learning for User Privacy and Data Confidentiality @ ICML 2020
arXiv / Workshop Version / code

Research Experience

 
 
 
 
 

Research Intern

Adobe Research

April 2020 – July 2020 Banglore (Remote)
 
 
 
 
 

AI Research Intern

Wellnesys Tech.

December 2018 – December 2018 Bangalore
Researched on Human Activity Recognition using IMU sensors (wearable devices) and a pressure sensitive yoga mat

Research Projects

Contact for more details

 
 
 
 
 

Affective Language Modelling and Text Generation

Prof. Ashutosh Modi

January 2020 – Present IIT Kanpur
Developed a model capable of generating intended affect-driven sentences with grammatical correctness intact.
 
 
 
 
 

Predicting Web Browsing Patterns - A Recommendation System Extension

Prof. Nisheeth Srivastava

August 2019 – Present IIT Kanpur
Built meta representations for the URLs from the web browsing history, topic modelled these representations, and analysed browsing patterns.
 
 
 
 
 

Stochastic Algorithms for Minimax Optimization

Prof. Ketan Rajawat

August 2019 – November 2019 IIT Kanpur
  1. Theoretically proved uncertainty of convergence of Stochastic variant of Accelerated GD (SAGD) for minimax problems, even though various practical implementations worked better than the deterministic version in terms of time complexity on large size problems.
  2. Proposed and proved convergence of 2 stochastic algorithms for general non-smooth minimax problems (convex-concave, strongly convex-concave). Further, formulated stochastic version of Efficient Algorithms for Smooth Minimax Optimization for non-convex-concave problems.
 
 
 
 
 

Brain Computer Interface - Predicted Mental Thought Processes using Neural Oscillations

May 2017 – July 2017 IIT Kanpur
Classified and recorded brain signals for 4 mental tasks: Relax, Calculate, Count, visualise 3-D object, as data using EEG Headset. Predicted with 88% accuracy in real time by using Ensemble Subspace KNN.

Posts

Find the blogs on my Medium space

Getting through Adobe India WIT Scholarship

Check it out on medium!

Scientific Discovery and it’s Logic — A review

Find the article here on Medium.

Social Limits on Freedom of Knowledge, based on ”The Dead Past” by Isaac Asimov

At the end of the story, the four main characters are in agreement with regard to the dangerous impact of the chronoscope on society. Does this final agreement among these characters mean that all debates related to knowledge within the story are clearly resolved at the end of the story?

Work Culture Differences and its Relation with Nation’s Growth: India and Japan

Work culture of a place encloses its history and the beliefs which are developed over time. Further, it also relates with the way the society interacts, preferred personalities of individual and society, social orders and hierarchies, etc.

Can an AI Acquire Human Personality?

Find the article here on Medium. I wrote this long time back, but published later on medium.

Contact

Visitors since Oct 2020,
Flag Counter