Welcome to COMSM0045. The unit introduces the students to deep architectures for learning linear and non-linear transformations of big data towards tasks such as classification and regression. The unit paves the path from understanding the fundamentals of convolutional and recurrent neural networks through to training and optimisation as well as evaluation of learnt outcomes. The unit's approach is hands-on, focusing on the 'how-to' while covering the basic theoretical foundations. For further general information, see the syllabus for the unit.
UPDATE - 29/08/2024 Welcome new students for the 24/25 year! PLEASE NOTE: lecture content will be updated, slides below are placeholder and may change until the lecture If you have any questions, head to the unit teams.Michael Wray (MW) | Unit Director |
Tilo Burghardt (TB) |
Jacob Chalk (JC), Omar Emara (OE), Rhdori Guerrier (RG), Sam Pollard (SP), Saptarshi Sinha (SS), Siddhant Bansal (SB), Zhifan Zhu (ZZ)
Wks | Monday 14:00-16:00 | Tuesday 09:00-12:00 | Labs |
1 |
16/09/2024 - 14:00 - Queens BLDG 1.07 Wk1 - LECTURE 1 INTRODUCTION TO THE UNIT intro PDF BASICS OF ARTIFICIAL NEURAL NETWORKS (Queens Building 1.07, in-person) (Introduction, Neural Networks, Perceptron, Cost Functions, Gradient Descent, Delta Rule, Deep Networks) PDF Slides, Recording |
17/09/2024 - 09:00 - MVB 1.15 Wk1 - LECTURE 2 TOWARDS TRAINING DEEP FORWARD NETWORKS (MVB 1.15, in-person + lecture recap) (Network Representation, Computational Graphs, Reverse Auto-Differentiation) PDF Slides, Extra Recap Recording Lecture 2 Refresher (first part of video) |
GETTING STARTED: Register Individually on BlueCrystal4 (details see below) RECAP WORKSHEETS: -Lab0 - Python (Homework) |
2 |
23/09/2024 - 14:00 - Queens BLDG 1.07 Wk2 - LECTURE 3 BACKPROPAGATION ALGORITHM (Queens Building 1.07, in-person + recorded lecture) (The Backpropagation Algorithm in Full Detail, Activation Functions) PDF Slides Extra Recap Recording (second part of video) Wk2 - LECTURE 4 OPTIMISATION TECHNIQUES (Queens Building 1.07, in-person + recorded lecture) (Stochastic Gradient Descent, Nesterov Momentum, RMSProp, Newton's Method, AdaGrad, Adam, Saddle Points) PDF Slides Extra Recap Recording |
24/09/2024 - 09:00 - MVB 1.15 - PRACTICAL 1 (Slides, Recording) Your first fully connected layer gradient descent stochastic gradient descent |
24/09/2024, (MVB 1.15) - 3hrs -BC4 Setup Lab 1 - Training your first Deep Neural Network |
3 |
30/09/2024 - 09:00 - Queens BLDG 1.07 Wk3 - LECTURE 5 COST FUNCTIONS, REGULARISATION AND DEPTH (Queens Building 1.07, in-person + recorded lecture) (SoftMax, Cross Entropy, L1 and L2 Regularisation, DropOut, DropConnect, Depth Considerations) PDF Slides Extra Recap Recording |
01/10/2024 - 09:00 - MVB 1.15 - PRACTICAL 2 (Slides, Recording). Your first convolutional connected layer |
01/10/2024, (MVB 1.15) - 3hr Lab 2 - Your First Convolutional Connected Network |
Wk3 - LECTURE 6 CONVOLUTIONAL NEURAL NETWORKS (Queens Building 1.07, in-person + recorded lecture) (sharing parameters, conv layers, pooling, CNN architectures) (Slides, Recording) |
|||
4 | - |
08/10/2024 - 09:00 - MVB 1.15 -
PRACTICAL 3 (Slides, Video Recording). Error rate monitoring (training/validation/testing) Batch-based training Learning rate Weight Freezing Batch normalisation Parameter intialisation |
08/10/2024, (MVB 1.15) - 3hr Lab 3 - Hyperparameters |
5 |
14/10/2024 - 09:00 - Queens BLDG 1.07 Wk5 - LECTURE 7 RECURRENT and RELATIONAL NEURAL NETWORKS (Queens Building 1.07, in-person + Recorded) (RNN, encoder-decoder, Transformers) (Slides, Video Pt. 1) Wk5 - LECTURE 8 (Queens Building 1.07, in-person + Recorded) GENERATIVE MODELS (Autoregressive models) (Slides, Video pt. 1, Video pt. 2) |
15/10/2024 - 9am - MVB 1.15 - PRACTICAL 4 (Slides). Data Augmentation Debugging strategies Dropout (Video Recording) |
15/10/2024, (MVB 1.15) - 3hr Lab 4 - Data Augmentation |
6 | READING WEEK - Mid Term for MAJOR unit students 25/10/2024 - MVB - 1.07 - 09:30-10:30 | ||
7 | - |
29/10/2024 - 09:00 - MVB 1.15 - PRACTICAL 5 (Slides). Transformers |
29/10/2024, (MVB 1.15) - 3hr Lab 5 - Transformers |
8 | - | 05/11/2024 - 09:00 - MVB 1.15 - Continuation Lab |
05/11/2024, (MVB 1.15) - 3hr |
9 | - | 12/11/2024, 10:00 [2 hours], (MVB 1.15) - CW Support Session |
- |
10 | - | 19/11/2024, 10:00 [2 hours], (MVB 1.15) - CW Support Session |
- |
11 | - | 26/11/2024, 10:00 [2 hours], (MVB 1.15) - CW Support Session |
- |
12 | - | 03/12/2024, MVB 1.15 - Exam Support Session | - |
13 | DECEMBER EXAMS - Final for MINOR unit students |
The coursework has been released, information can be found here
Note that the deadline is the 29th November at 13:00
Form for Sign-Ups. Even if you plan to work solo, please fill in this form.
If you need help, ask on the unit teams and/or come to the weekly CW support sessions.
Please note that you cannot take notes into the exam, but calculators are permitted.
All technical resources will be posted on the COMSM0045 ADL Github organisation. If you find any issues, please kindly raise an issue in the respective repository.
Recommended Reading:
Goodfellow et al (2016). Deep Learning. MIT Press
All students must apply online to register an account on BC4 for this unit. This also applies to students who already have accounts on BC4 for other units (e.g. HPC), in this case you must register again using the instructions below.
Note that it takes up to 48 hours to enable your account on BC4.