Laboratory work

Topic

The topic for the lab is deep learning. The goal of the lab is

• to learn how to build and train a neural network,
• to learn how to improve the neural network model and its training, and
• to get a glimpse of a state-of-the-art software library (PyTorch) for deep learning.

We will look at two applications with image classification, namely

• classification of hand-written digits by training and using a neural network, and
• classification of real world images using a pre-trained deep neural network.

Format

The lab is scheduled for 4 hours and done in groups. There are different sessions scheduled, each group should attend one of these. The lab is mandatory and examination will be done during these 4 hours, no report needs to be written.

Solutions to the preparatory exercises must be handed in before the lab session. The solutions to the preparatory exercise should be submitted here in Studium as a pdf.

Instructions

All instructions are in this file: instructions.pdf
Please read them carefully!

Additional material:

• introduction.ipynb
• code.zip

You can either choose to do the lab in your web browser via Google Colab or locally on your computer. If you choose the last option, make sure to have all software (notably PyTorch) installed before the lab session (see the instructions for details).

Use this document (Links to an external site.) for queuing for help during the lab.

Important dates

The deadlines for the lab are as follows:

(submission deadlines are one minute before midnight, i.e., 23.59)

Lab registration

You form groups when signing up.

Every group has to register for exactly one of the following three lab sessions in the course calendar:

• Wednesday, 15th December, 13:15-17:00
• Thursday, 16th December, 8:15-12:00
• Thursday, 16th December, 13:15-17:00

You can find the slots via the "Find appointment" button.

Questions?

If you have any questions regarding the lab, primarily use the Discussion forum - Laboratory work.