I am Maunil Vyas, a graduate student in Computer Science at Arizona State University (ASU). I believe that for the foreseeable future, automation and Artificial Intelligence will be the biggest contributors to the betterment of society. Therefore, I work in the field of Machine Learning (ML) and Deep Learning . My primary area of interest lies in the applied and theoretical aspects of ML. Specifically, in the theoretical aspect, I desire to understand the learning mechanism of the human brain, so that I can build more powerful Neural Network models. On the other hand in the applied aspect, I wish to enable the AI-based technology on a day to day life, therefore, I work in the field of Computer Vision (CV) and Natural Language Processing (NLP) . Apart from ML, I mostly work on improving my Software Development skills to improve as a developer.
To fulfil my research interest I am currently working in two research labs at ASU.
Research areas (Deep Learning)
1) Zero-Shot Learning (Thesis work) : Recently Submitted my Novel generative model to ECCV 2020.
2) Privacy and Fairness in Machine Learning: Working on generating fair data representation using GANs to preserve fairness as well as data privacy. Work highlights
Currently I am also working at American Express as a part-time Data Scientist.
I will be graduating in May 2020 and seeking full-time opportunities, Specifically in Machine Learning and Data Science fields. I am also open to Software Development.
Please feel free to contact at +1-4403816682 or email@example.com
E-commerce customer retention system: Built a complete internal recommendation system that helps to see the performance of various ML models on daily basis using Kibana, Elastic Search, Pyspark and Hadoop. Dealing with the different aspects of the ML, related to data pipeline building, feature engineering, modelling and performance reporting everyday.
Tech : Python, Pyspark, Hadoop, Elastic Search
Zero-Shot Learning for Visual Object Recognition: Developed a novel Generative model to perform Zero-Shot recognition for visual object recognition. Introduced a unique loss that helps to address the overfitting concern of the generative zero-shot model. The work is curently under review at European Conference on Computer Vision (ECCV) 2020 The proposed model leverages the semantic information in a form of either clean attributes or Wikipedia based text to perform the knowledge transfer.
Tech : Tensorflow, Pytorch, Python
Fair Data Generation: Developed Generative Adversarial Network-based systems to remove the inherent bias from the data set. Specifically, worked with the UCI Adult data set, Further details are discussed in (See my name in the acknowledgement section), Learning Generative Adversarial Representations under Fairness and Censoring Constraints
Tech : Tensorflow, Pytorch, Python
Data Collection Tools: Implemented web tools to help people label Audio (Speech) data online using Mechanical Turk and Amazon
Tech : Python Flask, Java script, HTMLTranscription Cleaning: Worked on Auto Encoders based models to denoise the transcript data
Tech : Pytorch, Python
Worked as Instructional Aide for the courses MAT 142: College Mathematics and MAT 266: Calculus for Engineers II.
Being a teaching assistant in the course of Introduction to Computer Programming and Computer Networks, my primary responsibility was to support the professor to conduct the course smoothly. Mainly, I was accountable for taking recitation classes and doing the exam evaluation.
ML Based Spectrum Sensing Scheme: I designed an ML-based spectrum sensing scheme for cognitive radio which significantly outperformed the conventional approaches, and published it in PIMRC 2017;
Tech : Python, Keras, MATLAB , USRPs, Disconn antenna
Under review at European Conference on Computer Vision (ECCV) 2020 a novel Generative model to address the Zero-Shot
Learning for object recognition. - First Author
Artificial Neural Network Based Hybrid Spectrum Sensing Scheme for Cognitive Radio
M. R Vyas , D. K Patel, M. López-Benítez
Proceedings of the 28th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communication (PIMRC 2017), Montreal, Quebec, Canada, October 8-13, pp xxx-xxx. (13 Citations, as of March 2020)
Developed a Generative Adversarial Model having a novel Alpha loss to address the fairness concerns in Deep Learning. Specifically, worked on UCI Adult dataset.
Tech : Python, TensorFlow
Created a web-based visualization system for Hotel Recommendation. I worked on the text mining part of the system. Implemented a sentiment visualization from user comments, developed a word cloud from the sentiments of the comment words. Designed a comment similarity system to find the top comments using the Natural language inference idea.
computer vision application for the surveillance using Drone. The system guides the drone to follow a person, who matches the desired soft features such as color of the cloths. Used Mobile Net for person detection and ResNet for the color classification. The KCF tracker is utilised to track the individual in the video frame. Tech: Python, Pytorch, OpenCV
Tech : Python, Pytorch, OpenCV
Modified the attention model of the Decomposable Attention Model for Natural Language Inference Model.
Tech : Python, Pytorch
Implemented a paper titled : Face recognition using reinforcement learning .
Tech : Python
Improved my C++ skills by implementing various features for a compiler. Using the basic language grammar, implemented a predictive recursive parsing for Syntax Check, a type checker for Semantic Check and singly link list for program execution.
Tech : C++
Represented the weight learning problem of the Neural Network as a constraint satisfaction problem. Using augmented Lagrangian tried to solve the optimization problem. Successfully trained NN for two-class classification without gradient based back propagation. Python, Theoretical Machine Learning, Mathematical Optimization
Tech : Python, Theoretical Machine Learning, Mathematical Optimization
A funded project, aimed to build a music synthesizer for Indian classical instrument Harmonium with a facility of 22 Shruti (“Notes”).
Tech : Python, Scilab, MATLAB, Music-Theory, C++
Generated artificial images using GANs. Worked on techniques to make the Generator network faster using Probabilistic Principal Component Analysis (PPCA).
Tech : Python, Tensorflow, MATLAB
Apart from being an engineer, I enjoy most of my time being indoors. Playing my Piano and Guitar in front of a huge crowd is my lifelong dream. I also manage to sing well thus, my passion for music in playing and singing lead me in the world of music recording. I have a small recording studio. Check out my YouTube Channel
After music, I prefer to watch sports, spcifically Cricket. I am MSD Fan.
Last but not least, I enjoy spending time with my family.