agisga/CS231n_2018_solutions

My solutions to the Stanford CS231n (Spring 2018) assignments

Stanford CS231n Convolutional Neural Networks for Visual Recognition

• Course website: https://cs231n.github.io/
• Assignment 1 (Spring 2018): https://cs231n.github.io/assignments2018/assignment1/
• Assignment 2 (Spring 2018): https://cs231n.github.io/assignments2018/assignment2/
• Assignment 3 (Spring 2018): https://cs231n.github.io/assignments2018/assignment3/

This repository contains my notes & solutions to the assignments.

Please note that I was not enrolled for the course and my solution was not submitted, checked or graded.

Docker

• I run the assignments using a Docker container.
• For problems that don't require TensorFlow or PyTorch I use a basic miniconda-based container (based on the continuumio/miniconda3 Docker container), which is set up in a way similar to what I have described here.

  • To run the container:

    docker run -p 9999:8888 --name CS231n -v ~/github/my_CS231n/:/app/data cs231n
    

    where cs231n is the name of my Docker image.

  • To restart the container after it has shut down:

    docker start -ia CS231n
    

    where CS231n is the name of my Docker container.

• For problems that use PyTorch I either use this Dockerfile locally, or work on AWS without Docker (see below).

AWS

• To run the more computationally heavy stuff that uses TensorFlow or PyTorch, I use AWS spot instances initialized with Amazon's "Deep Learning AMI (Ubuntu)" image. Here is a description of my workflow (under the section "AWS Deep Learning AMI").