Block 5: Neural networks
Suggested work for this block
4 hours neural network lecture, 2 hour reading and video watching and 2 hours peer-review of mini-project, 4 hour preparation for neural network lab, 2 hours pen-and-paper turorials, 2-8 hours on revising mini-project based on peer-review comments
Lecture F7: Neural networks and deep neural networks
Pedagogical presentation of what a neural network is. The videos also serve as a good preparation of the laboratory work since they work with the same example and data as you will work with there. Links to an external site.
The Neural Network Playground Links to an external site. is a very good tool to get a sense of what neural network models can look like.
Lecture F7 on Friday Nov 25th @ 13.15
Recommended reading: Ch. 6.1 + 3.2
Lecture F8: Convolutional neural network models and training neural networks
The following video is a continuation of the video above where they explain who we train neural networks from training data.
Lecture F8 on Friday Dec 1st @ 10.15
Recommended reading: Chapters 6.2-6.3 and 5.3-5.5. (You can skip "Backpropagation" in section 6.2.)
Tutorial 10 (pen & paper)
Start working on this before you start with preparing the lab.
The exercise sheet is available here Links to an external site..
Recommended problems: 10.1, 10.2, 10.3, 10.4 (all)
Feel free to ask questions in the discussion section
Compulsory lab session: Neural networks
Connected to the topic of neural networks there is a compulsory lab session. Remember to do and submit answers to the preparatory exercises before the lab session. You find more information about the format, the material and how to sign up here.
Feel free to ask questions in the discussion section.
Want to learn more?
Deeper understanding
In "Backpropagation" in Section 6.2 and Section 6.A in the SML book covers the backpropagation algorithms and its derivation and Section 6.4 introduces one popular regularization for neural networks called dropout. Both of are not covered in the lectures (and is outside the scope of the course) but are highly interesting!
Online courses on deep learning
There are plenty of resources online if you want to learn more about deep learning. If you want to spend some more time on the topic, the online course platform Coursera offers a popular series of courses on the topic which you can find here Links to an external site.. These course modules cover a broad set of deep learning models and also hand-on techniques for developing supervised machine learning models in general.