Introduction to Deep Learning

Assignment 1

Implement and train a deep model on a MNIST image classification task.

Details: Assignment 1

Preparing the MNIST dataset

1. Download the raw MNIST files from Yann LeCun’s website
2. Unzip the data files.
3. Copy the files in to course_intro-to-dl/data/MNIST
4. Download the coversion.py script.
5. Run python conversions.py -c -n

Model

Multi-layer perceptron

Accuracy

Accuracy: 89.2188%

Assignment 2

Basic Visualization on a MNIST image classification task.

Details: Assignment 2

Preparing the MNIST dataset

1. Follow this to import data into google colaboratory

Setup Tensorboard

2. Setup tensorboard in google colab.

Assignment 3

Create a model for the MNIST dataset using convolutional neural networks (CNN).

Details: Assignment 3

Preparing the MNIST and Fashion-MNIST dataset

1. Download the raw MNIST files from Yann LeCun’s website
2. Download the raw Fashion-MNIST files from Fashion-MNIST
3. Unzip the data files.
4. Copy the files in to course_intro-to-dl/data/MNIST
5. Download the coversion.py script.
6. Run python conversions.py -c -n

Model

Convolutional Neural Networks

Accuracy

MNIST Accuracy: 97.3600%

Fashion MNIST Accuracy: 83.1400%

Assignment 4

Using Estimator for

Details: Assignment 4

Preparing the MNIST and Fashion-MNIST dataset

1. Download the raw MNIST files from Yann LeCun’s website
2. Unzip the data files.
3. Copy the files in to course_intro-to-dl/data/MNIST
4. Download the coversion.py script.
5. Run python conversions.py -c -n

Model

Convolutional Neural Networks

STOPPED, because I won't be working in Tensorflow anymore. PyTorch is way better and more intuitive.