-
Notifications
You must be signed in to change notification settings - Fork 3
/
index_2020.xhtml
239 lines (202 loc) · 14 KB
/
index_2020.xhtml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN" "http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en-gb" xml:lang="en-gb">
<head>
<!--nomodify-->
<meta name="style" content="cs" />
<title>University of Bristol - Computer Science Department - COMSM0045 - Applied Deep Learning</title>
<style>
table { border-spacing: 2px; border-collapse: separate; }
td { margin: 1px; padding: 5px; }
table.blank {vertical-align: top; text-align: left; background: #f1f1f1;}
td.blank {vertical-align: top; text-align: center; background: #cccccc;}
tr.blank {vertical-align: top; text-align: left; background: #cccccc;}
td.dima {vertical-align: top; text-align: left; background: #FFE5B4;}
td.tilo {vertical-align: top; text-align: left; background: #ccffc1;}
td.assess {vertical-align: top; text-align: left; background: #ffaaaa;}
td.labs {vertical-align: top; text-align: left; background: #bbbbee;}
</style>
</head>
<body>
<div id="wrap">
<a href="https://www.bris.ac.uk/unit-programme-catalogue/UnitDetails.jsa?unitCode=COMSM0045" target="_blank">UNIT INFO</a>
<h1>COMSM0045 - Applied Deep Learning</h1>
<img src="comsm0018.jpg" alt="ADL Banner" width="900"></img>
<link rel="stylesheet" href="simple.css" />
<a id="info"></a>
<hr/>
<h2>Unit Information</h2>
<p>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
<a href="https://www.bris.ac.uk/unit-programme-catalogue/UnitDetails.jsa?unitCode=COMSM0045" target="_blank">
the syllabus for the unit</a>.
</p>
<hr/>
<h2> Staff</h2>
<table>
<tr><td class="dima"><a href="http://dimadamen.github.io" target="_blank">Dima Damen (DD)</a></td><td><b>Unit Director</b></td></tr>
<tr><td class="tilo"><a href="http://www.cs.bris.ac.uk/~burghard" target="_blank">Tilo Burghardt (TB)</a></td><td></td></tr>
</table>
<hr/>
<h2> Teaching Assistants</h2>
<p>Hazel Doughty (HD), Will Price (WP), Evangelos Kazakos (EK), Jonathan Munro (JoM), Jian Ma (JiM), Xinyu Yang (XY), Dan Whettam (DW), Adriano Fragomeni (AF)</p>
<a id="materials"></a>
<hr/>
<h2>Unit Materials</h2>
<table class="blank">
<tr class="blank">
<td><i>Wks</i></td> <td><i>Pre-Recorded Lectures</i></td> <td><i>Friday Synch Session</i><br/>10am-1pm</td> <td><i>Labs</i></td>
</tr>
<tr>
<td class="blank">0</td>
<td class="dima">
Wk0<br/>
<b><a href="https://web.microsoftstream.com/video/a8a81ee6-a8da-42cf-87b9-28e996dc52af">INTRODUCTION TO COMSM0045 (Video)</a></b>
</td>
<td></td>
<td><b>GETTING STARTED:</b><br/>
<hr/><a href="#bc4" target="_blank">Register Individually on BlueCrystal4<br/>(details see below)</a>
</td> </tr>
<tr>
<td class="blank">1</td>
<td class="blank"></td>
<td class="tilo">
Wk1 - LECTURE 1<br/>
<b>BASICS OF ARTIFICIAL NEURAL NETWORKS</b><br/>(Live on Teams at 10am 09/10/20, <a target="_blank" href="Slides/COMSM0045_01.pdf">Slides</a>)<br/>
(Introduction, Neural Networks, Perceptron, Cost Functions, Gradient Descent, Delta Rule, Deep Networks)<br/><hr/>
Wk1 - LECTURE 2<br/>
<b>TOWARDS TRAINING DEEP FORWARD NETWORKS</b><br/>(Live on Teams at 11am 09/10/20, <a target="_blank" href="Slides/COMSM0045_02.pdf">Slides</a>)<br/>
(Network Representation, Computational Graphs, Reverse Auto-Differentiation)
</td>
<td>(no scheduled lab for week 1)<hr/> <b>RECAP WORKSHEETS:</b><br/>-Convolutions (Homework)<br/>-<a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets">Lab0 - Python (Homework)</a>
</td> </tr> <tr>
<td class="blank">2</td>
<td class="tilo">
Wk2- LECTURE 3<br/>
<b>BACKPROPAGATION ALGORITHM</b><br/>(<a target="_blank" href="https://mediasite.bris.ac.uk/Mediasite/Play/4db6e3234d314a1aa2935d71c02a0e071d">Video</a>, <a target="_blank" href="Slides/COMSM0045_03.pdf">Slides</a>)<br/>
(The Backpropagation Algorithm in Full Detail, Activation Functions)<br/><hr/>
Wk2 - LECTURE 4<br/>
<b>OPTIMISATION TECHNIQUES</b><br/>(<a target="_blank" href="https://mediasite.bris.ac.uk/Mediasite/Play/aa25459da02e430a8c99e3cfd5082a371d">Video</a>, <a target="_blank" href="Slides/COMSM0045_04.pdf">Slides</a>)<br/>
(Stochastic Gradient Descent, Nesterov Momentum, RMSProp, Newton's Method, AdaGrad, Adam, Saddle Points)<hr/>
<!--LECTURE 6<br/>
<b>REGULARISATION AND AUGMENTATION</b><br/>
(L1 and L2 Weight Decay, Dropout, Data Augmentation)<hr/>-->
<!--LECTURE 7<br/><b>SCALE AND DEPTH</b><br/>
(Why deep?, How well does it scale?, Limitations?-->
</td>
<td class="dima">
16/10/20,10am,Online - <b>PRACTICAL 1</b> <a href="Slides/COMSM0045_Practical1-Lab1-2020.pdf">(Slides)</a>, <a href="https://web.microsoftstream.com/video/f7a0df6b-5255-4252-aba2-cb6b6dcf95b4">(Video)</a><br/>
Your first fully connected layer<br/>
gradient descent<br/>
stochastic gradient descent
</td>
<td class="labs">16/10/20, Online - 3hrs<br/>-<a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets/blob/master/lab-1-dnns/bc4-setup.ipynb">BC4 Setup</a>
<hr/>
<a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets"><b>Lab 1</b> - Training your first Deep Neural Network</a>
</td>
</tr>
<tr>
<td class="blank" rowspan="2">3</td>
<td class="tilo">Wk3 - LECTURE 5<br/>
<b>COST FUNCTIONS, REGULARISATION AND DEPTH</b><br/>(<a target="_blank" href="https://mediasite.bris.ac.uk/Mediasite/Play/9fd25a71068443d18c1689759d3e79f31d">Video</a>, <a target="_blank" href="Slides/COMSM0045_05.pdf">Slides</a>)<br/>
(SoftMax, Cross Entropy, Hingeloss, L1 and L2 Regularisation, DropOut, DropConnect, Depth Considerations)<br/>
</td>
<td class="dima" rowspan="2">
23/10/20,10am,Online - <b>PRACTICAL 2</b> <a href="Slides/COMSM0045_Practical-Lab2-2020.pdf">(Slides)</a>, <a href="https://web.microsoftstream.com/video/2c73f334-9ee9-4a28-9b52-e310d0820eb0">(Video)</a><br/>
Your first convolutional connected layer
</td>
<td class="labs" rowspan="2">23/10/20, Online - 3hr<br/> <a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets"><b>Lab 2</b> - Your First Convolutional Connected Network</a></td>
</tr>
<tr>
<td class="dima">
Wk3 - LECTURE 6<br/><b>CONVOLUTIONAL NEURAL NETWORKS</b><br/>
(<a target="_blank" href="https://web.microsoftstream.com/video/84b02183-0723-4240-8b6c-aa4aeb96e86a">Video Part 1</a>, <a target="_blank" href="https://web.microsoftstream.com/video/aef7d812-c7c4-4ade-83a7-61496c4ab3a5">Video Part 2</a>, <a target="_blank" href="Slides/COMSM0045-convolutional-networks-Part1.pdf">Slides Part 1</a>, <a href="Slides/COMSM0045-convolutional-networks-Part2.pdf">Slides Part 2</a>)<br/>
(sharing parameters, conv layers, pooling, CNN architectures)
</td>
</tr>
<tr>
<td class="blank">4</td> <td class="blank">-</td>
<td class="dima">
30/10/20,10am,Online -
<b>PRACTICAL 3</b> <a href="Slides/COMSM0045_Practical_Lab3-2020-handout.pdf">(Slides)</a>, <a href="https://web.microsoftstream.com/video/d35b7cf2-943d-4872-8f42-45434a7fca61">(Video)</a><br/>
Error rate monitoring (training/validation/testing)<br/>
Batch-based training<br/>
Learning rate<br/>
Batch normalisation<br/>
Parameter intialisation</td>
<!-- <a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0045_2018/content/COMSM0045_Practical2_handout.pdf" target="_blank">slides</a>-->
<td class="labs">30/10/20, Online - 3hr<br/><br/><b><a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets">Lab 3 - Hyperparameters</a></b></td> </tr>
<tr> <td class="blank">5</td> <td class="blank">-</td> <td class="dima">6/11/20, 10am, Online<br/><b>Continuation Lab</b><br/>
</td> <td class="labs">6/11/20, Online - 3hr<br/><br/><a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets"><b>Lab</b><br/>Continuation</a></td></tr>
<tr>
<td class="blank">6</td>
<td class="blank">-</td>
<td class="dima">
13/11/20, 11am, Online - <b>Practical 4 Intro</b> <a href="Slides/COMSM0045_Practical_Lab4-2020-handout.pdf">(Slides)</a> <a href="https://web.microsoftstream.com/video/ca8c5708-0617-4557-9d7d-1d677e756b7f">(Video)</a><br/><hr/>
<b>PRACTICAL 4</b><br/> Data Augmentation<br/>Debugging strategies<br/>
Dropout
</td>
<td class="labs">13/11/20, Online - 3hr<br/><a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets"><b>Lab 4</b><br/>Data Augmentation</a></td>
</tr>
<tr>
<td class="blank">7</td>
<td class="dima">Wk7 - LECTURE 7 <br/><b>RECURRENT NEURAL NETWORKS</b><br/>
(temporal dependencies, RNN, bi-directional RNNs, encoder-decoder, LSTM, gated RNN)<br/>
<a href="./Slides/COMSM0045-recurrent-networks.pdf">(Slides)</a><a href="https://web.microsoftstream.com/video/6faf0a25-537e-4819-afb7-17d4cea15583">(Video)</a>
</td>
<td class="dima">20/11/20, 10am, Online<br/><b>Continuation Lab + First CW Lab</b><br/></td>
<td class="labs">20/11/20, Online - 3hr<br/><br/><a href="https://github.com/COMSM0045-Applied-Deep-Learning/labsheets"><b>Lab</b><br/>Continuation + Project Start</a></td>
</tr>
<tr> <td class="blank">8</td> <td class="blank">-</td> <td class="dima">27/11/20, 10am [1 hour], Online - <b>CW Q/A</b></td>
<td class="blank">-</td></tr>
<tr>
<td class="blank">9</td> <td class="blank">-</td> <td class="dima">4/12/20, 10am [1 hour], Online - <b>CW Q/A</b></td>
<td class="blank">-</td>
</tr>
<tr>
<td class="blank">10</td> <td class="blank" colspan="3">CW Deadline Fri 11 Dec 13:00 (Blackboard Submission)</td> </tr>
<tr> <td class="blank">11</td> <td class="blank">-</td> <td class="dima">18/12/20, 10am [1 hour], Online - <b>Exam Q/A</b></td><td class="blank">-</td></tr>
</table>
<hr/>
<h2>Assessment Details</h2>
All students in the unit are requested to study the paper:
<b>Pan et al (2016). Shallow and Deep Convolutional Networks for Saliency Prediction.</b> IEEE Conference on Computer Vision and Pattern Recognition (CVPR). <a href="https://arxiv.org/abs/1603.00845">Available on ArXiv</a>
<p>Coursework specs are now available at: <a href="./Assessment/COMSM0045_Coursework_2020.pdf">COMSM0045-COURSEWORK-SPECS-2020</a></p>
<ul>
<li><b>Coursework: </b> You are requested to re-implement the paper above, and provide your code as well as a final report. Coursework completed in groups (up to 3)</li>
<li><b>Exam: </b> In addition to the content of the unit, students are expected to read and study the paper above. 25% of the exam will involve this paper, and its relationship to topics you have studied in the unit.</li>
</ul>
<p> </p>
<hr/>
<h2>Assessment Details - Exam</h2>
<p>Open book 2 hours exam in January (online). See note above on additional reading for exam. </p>
<hr/>
<h2>Github</h2>
<p>All technical resources will be posted on the
<a href="https://github.com/COMSM0045-Applied-Deep-Learning" target="_blank">COMSM0045 ADL Github organisation</a>. If you find any issues, please kindly raise an issue in the respective repository.
</p>
<hr/>
<h2>Textbook</h2>
<p>Recommended Reading:<br/><a href="http://www.deeplearningbook.org/" target="_blank">Goodfellow et al (2016). Deep Learning. MIT Press</a></p>
<hr/>
<h2><a id="bc4">Blue Crystal 4 Registration [only applicable for Bristol undergraduate students with corresponding email]</a></h2>
<p>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.</p>
<ol>
<li>Click on: <a href="https://www.acrc.bris.ac.uk/login-area/apply.cgi" target="_blank">https://www.acrc.bris.ac.uk/login-area/apply.cgi</a></li>
<li>Enter your personal details</li>
<li>Choose: "Join an existing project"</li>
<li>Enter project code: COSC020582</li>
<li>Keep Preferred log-in shell as bash</li>
<li>Do not provide any additional information</li>
</ol>
<p>Note that it takes up to 48 hours to enable your account on BC4.</p>
</div>
</body>
</html>