Chris Pollett > Students > More

    Print View

    [Bio]

    [Blog]

    [CS 297 Proposal]

    [Del 1 - Write LaTex for Famous Mathematical Equations]

    [Del 2 - Exploring TensorFlow with MNIST Dataset]

    [Del 3 - Exploring Data Generation Approaches]

    [Del 4 - Decide the Neural Network Architecture]

    [CS 297 Report-PDF]

    [CS 298 Proposal]

    [CS 298 Report-PDF]

    [CS 298 Slides-PDF]





































Project Blog

Summary of the Meeting on May 8, 2018

  • Showed all the slides which I had prepared for the defense.
  • Dr. Pollett reviewed the slides and suggested changes. Dr. Pollett also gave me some suggestions on how to present and defend the project work.

Summary of the Meeting on May 1, 2018

  • Showed complete first draft of the report
  • Dr. Pollett reviewed the report, suggested changes and then approved the report for submission to the department.

Summary of the Meeting on Apr. 24, 2018

  • Showed progress on report.
  • Dr. Pollett reviewed experiements part of the report.

Summary of the Meeting on Apr. 17, 2018

  • Showed progress on report. Dr. Pollett approved the structure of report which I had proposed
  • Dr. Pollett reviewed implementation part of the report.

Summary of the Meeting on Apr. 10, 2018

  • Showed progress on matrix operations on matrices of 2x2 size
  • Dr. Pollett suggested me to start working on the report.

Summary of the Meeting on Apr. 3, 2018

  • Showed progress(working model for equations involving simple summation, integration and mathematical equations) on the deliverable involving complex mathematical equations
  • Dr. Pollett suggested me to train model for equations involving matrix opeations and stacked expressions in terms of power.

Summary of the Meeting on Mar. 20, 2018

  • Started working on deliverable involving complex mathematical equations.
  • We decided to explore the depth till which CNN can recognize characters

Summary of the Meeting on Mar. 13, 2018

  • Showed progress on second deliverable to Dr. Pollett.
  • Model was able to recongnize equations involving operations such as addition, subtraction, multiplication and exponenents with the length of 5 characters.

Summary of the Meeting on Mar. 6, 2018

  • Discussed possible ways to experiment to complete second deliberable with Dr. Pollett.

Summary of the Meeting on Feb. 27, 2018

  • Discussed my progress on the first deliverable with Dr. Pollett.
  • Shown results after bugs were fixed to make model work with accuracy higher than 90%
  • Dr. Pollett's suggested me to train model for alphabets and numbers

Summary of the Meeting on Feb. 20, 2018

  • Discussed my progress on the first deliverable with Dr. Pollett.
  • Dr. Pollett's suggested me to make model as close as LeNet and experiment with the different data sizes

Summary of the Meeting on Feb. 13, 2018

  • Discussed my progress on the first deliverable with Dr. Pollett.
  • Dr. Pollett's suggested me to change the data size from 30k to 200k and to experiment with the regularization methods

Summary of the Meeting on Feb. 6, 2018

  • Discussed CS 298 proposal with Dr. Pollett and got feedback on it.
  • Discussed the first deliverable. The first deliverable is to predict the first character of equations.

Spring 2018

Summary of the Meeting on Dec. 13, 2017

I showed the second draft of the 297 report and collected Dr. Pollett's feedback on it.

Summary of the Meeting on Dec. 5, 2017

I showed the first draft of the 297 report and collected Dr. Pollett's feedback on it.

Summary of the Meeting on Nov. 28, 2017

I presented the architecture diagram of the proposed solution to Dr. Pollett. We discussed the CS 297 report which is the last deliverable.

Summary of the Meeting on Nov. 21, 2017

Meeting canceled because CNN homework was due. This homework was related to my CS 297 problem statement.

Summary of the Meeting on Nov 14, 2017

I demoed the code to classify the equations based on the symbols in it. Professor advised me to explore the approaches and come up with an architecture for the network.

Summary of the Meeting on Nov 7, 2017

I showed generated data. We discussed how to integrate CNN and RNN.

Summary of the Meeting on Oct. 24, 2017

I presented my understanding of RNN and LSTM to Dr. Pollett. We discussed next two deliverables.
  • Deliverable 3: Data generation approaches
  • Deliverable 4: Architecture of the network

Summary of the Meeting on Oct. 17, 2017

Meeting cancelled because I was sick.

Summary of the Meeting on Oct. 10, 2017

I presented my understanding of CNN to Dr. Pollett. We discussed tasks for the following week.
  • Understand RNN and LSTM

Summary of the Meeting on Oct. 3, 2017

I presented my understanding of TensorFlow to Dr. Pollett. We discussed tasks for the following week.
  • Understand Chapter 9 of Deep Learing book by Goodfellow-et-al-2016

Summary of the Meeting on Sep. 26, 2017

I presented my understanding of "Show and Tell: A Neural Image Caption Generator" Paper to Dr. Pollett. We discussed tasks for the following week.
  • Work on the second deliverable which involved an understanding of Tensorflow with MNIST dataset

Summary of the Meeting on Sep. 19, 2017

I presented my understanding of "Image-to-Markup Generation with Coarse-to-Fine Attention" Paper to Dr. Pollett. We discussed tasks for the following week.
  • Read the paper by Oriol Vinyals, Alexander Toshev, Samy Bengio, Dumitru Erhan. Published in: 2015. Paper title: Show and Tell: A Neural Image Caption Generator

Summary of the Meeting on Sep. 12, 2017

Dr. Pollett reviews my first deliverable and discusses tasks for the following week.
  • Add matrix equation to the LaTex deliverable
  • Add recursive equation to the LaTex deliverable
  • Prepare powerpoint presentation on "Image-to-Markup Generation with Coarse-to-Fine Attention" paper

Summary of the Meeting on Sep. 5, 2017

Dr. Pollett reviews my proposal, and discusses tasks in the following week.
  • Fill out the bio, blog and proposal pages
  • Explore LaTex and write LaTex for some of the famous mathematical equations
  • Start reading the paper