Description

In this assignment you shall classify an image dataset, MNIST containing hand-written digits or Fashion-MNIST containing clothing, using neural network algorithms. You shall present and explain your code in an oral examination.

Development environment

See the Development environment page.

Submission instructions

See the Deadlines and Submissions page.

Requirements

• Write code for classifying either MINST or Fashion-MNIST dataset (both are available in most deep learning libraries)
• You can choose between the following machine learning libraries: Deeplearning4j (Java), Keras (Python) or Keras for R (R)
• Classify the dataset with both a standard neural network (no convolutional layers) and a convolutional neural network
• Each result shall be presented with accuracy score for the test set (the datasets are already split into train and test sets)

Getting started

Before starting with the assignment, watch lectures L06 and L07. L08 is also good to watch.

Here are some useful guides to get you started:

Scikit-learn:

• How to Develop a CNN for MNIST Handwritten Digit Classification
• MNIST digits classification dataset

Deeplearning4j:

• Quickstart with MNIST

Keras for R:

• Keras: Deep Learning in R

If you need help

Ask questions in the Slack channel or contact the main instructor to book an online meeting.