WORK AND PROJECT
-
Network Visualization
In order to improve our understanding of how CNN works, I worked on the assignment to create structures to generate Class Visualization, Saliency Maps, and Fooling Images. -
GAN
Vanilla GAN, Least Squares Distance GAN, and DCGAN implementation using Tensorflow -
Machine Translation
Built a deep neural network that functions as part of an end-to-end machine translation pipeline. The completed pipeline accepts English text as input and returns the French translation. Check the "IMPLEMENTATION" sections for my work. -
Simple Batch Normalization
This is a summarization of an attempt to obtain the analytical solution to back-propagation in a batch normalization layer and understand the mechanism underlying batch normalization. -
Feature Inversion and Style Transfer
Created a style transfer network and generated images using Tensorflow and an exsiting CNN network. Also, utilized the same structure and the gradients to generate an image back from the high level feature. -
Image Captioning
Built LSTM based network using numpy and created an image captioning network. -
LSTM with numpy
Built LSTM forward and back prop using numpy -
Transfer Learning with CNN
Built an algorithm to identify canine breed given an image of a dog. If given image of a human, the algorithm identifies a resembling dog breed. Applied transfer learning by utilizing an existing CNN deep neural network, Xception, and applying an average pooling layer and fully connected layer on the end to classify dog breeds. -
Vectorized Max Pooling with numpy
Implemented the forward and backward max pooling implementation layer in numpy. It started off with a naive implmentation then optimized into a vectorized impelmentation, which improved runtime by orders of magnitude. -
RNN Text Generation
Used the Sherlock Holmes novel as input corpus for English sequence generation. For code that runs my answers, you can check the Jupyter notebook (check "IMPLEMENTATION" for my contributions. Also, retyped all other code): https://github.com/baek-jinoo/aind2-rnn/blob/master/RNN_project.ipynb -
Logic and Planning Agent
Used logic and planning techniques to create an AI that finds the most efficient route to route cargo around the world to their respective destinations. This project used a combination of propositional logic and search along with A* heuristics to find optimal planning solutions. -
UC Berkeley Pacman Search
Finished the UC Berkeley Pacman search assignment. -
Constraint Satisfaction Problem
Implemented CSP algorithms to by implementing backwards and foward search. -
Isolation Game Agent
Built an agent to play isolation, the game. Utilized alpha-beta pruning and minimax algorithms. -
AVL-Tree
Objective-C implementation of AVL-Tree for insert and delete. Employed TDD -
Core Data Places
App built while learning iOS -
Asynchronous Robots
iOS application to learn about asynchronous code with run-loop sources