-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathindex_preEdits.xhtml
215 lines (187 loc) · 12.5 KB
/
index_preEdits.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
<!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 - COMSM0018 - 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.ole.bris.ac.uk/webapps/blackboard/execute/content/blankPage?cmd=view&content_id=_3080189_1&course_id=_226947_1" target="_blank">BLACKBOARD PAGE</a> |
<a href="http://www.bris.ac.uk/unit-programme-catalogue/UnitDetails.jsa?ayrCode=17%2F18&unitCode=COMSM0018" target="_blank">UNIT INFO</a> |
<a target="_blank" href="https://www.ole.bris.ac.uk/webapps/discussionboard/do/forum?action=list_threads&course_id=_226947_1&nav=discussion_board_entry&conf_id=_183420_1&forum_id=_148948_1">FORUM</a> |
<a href='https://wwwa.fen.bris.ac.uk/COMSM0018/' target="_blank">SAFE</a>
<h1>COMSM0018 - 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 COMSM0018. 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="http://www.bris.ac.uk/unit-programme-catalogue/UnitDetails.jsa?ayrCode=17/18&unitCode=COMSM0018" target="_blank">
the syllabus for the unit</a> and the <a href='https://wwwa.fen.bris.ac.uk/COMSM0018/' target="_blank">SAFE website</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>office 3.12 MVB. <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> office 3.42 MVB. </td></tr>
</table>
<hr/>
<h2> Teaching Assistants</h2>
<p>Hazel Doughty (HD), Davide Moltisanti (DM), Miguel Fortiz (MF), Michael Wray (MW), Will Price (WP), Will Andrew (WA), Evangelos Kazakos (EK), Jonathan Munro (JM)</p>
<a id="materials"></a>
<hr/>
<h2>Unit Materials</h2>
<table class="blank">
<tr class="blank">
<td><i>Wks</i></td> <td><i>Monday Sessions</i></td> <td><i>Friday Sessions</i></td> <td><i>Labs</i></td>
</tr>
<tr>
<td class="blank">1</td>
<td class="tilo">
01/10/18, 11am, QB.F101 - LECTURE 1<br/>
<b>BASICS OF ARTIFICIAL NEURAL NETWORKS</b><br/>
(Introduction, Neural Networks, Perceptron, Cost Functions, Gradient Descent, Delta Rule, Deep Networks)<hr/>
01/10/18, 1pm, QB.F101 - LECTURE 2<br/>
<b>TOWARDS TRAINING DEEP FORWARD NETWORKS</b><br/>
(Computational Graphs, Reverse Auto-Differentiation)
</td>
<td class="tilo">
05/10/18, 9am, QB.1.68 - LECTURE 3<br/>
<b>BACK-PROPAGATION</b><br/>
(The Backpropagation Algorithm)<hr/>
5/10/18, 11am, QB.1.68 - LECTURE 4<br/>
<b>OPTIMISATION TECHNIQUES</b><br/>
(SGD, Momentum, NAG, Newtons Method, Saddle Points, Adagrad, RMSprop, Adam)
</td>
<td>(no scheduled lab for week 1)<hr/> <b>GETTING STARTED:</b><br/>
<hr/><a href="#bc4" target="_blank">Register Individually on BlueCrystal4<br/>(details see below)</a>
<hr/><b>RECAP WORKSHEETS:</b><br/>-Convolutions (Homework)<br/>-<a href="https://github.com/COMSM0018-Applied-Deep-Learning/labsheets/blob/master/Lab_0_Python_Intro/0%20-%20Contents.ipynb">Python (Homework)</a>
</td> </tr> <tr>
<td class="blank">2</td>
<td class="tilo">
08/10/18, 1pm, QB.F101 - LECTURE 5<br/>
<b>ACTIVATION AND COST FUNCTIONS</b><br/>
(Key Activation Functions, Key Cost Functions)
</td>
<td class="tilo">
12/10/18, 9am, QB.1.68 - LECTURE 6<br/>
<b>REGULARISATION AND AUGMENTATION</b><br/>
(L1 and L2 Weight Decay, Dropout, Data Augmentation)<hr/>
12/10/18, 11am, QB.1.68 - LECTURE 7<br/><b>SCALE AND DEPTH</b><br/>
(Why deep?, How well does it scale?, Limitations?
</td>
<td class="labs">08/10/18, 12noon, QB.F101 - 1hr<br/>-<a href="https://github.com/COMSM0018-Applied-Deep-Learning/labsheets/blob/master/Lab_1_intro/Running%20TensorFlow.ipynb">BC4 Stress Test</a>
<hr/>
<b>Lab 1</b> - Introduction to TensorFlow
</td>
</tr>
<tr>
<td class="blank">3</td>
<td class="dima">15/10/18, 11am, QB.F101 - LECTURE 8<br/>
<b>CONVOLUTIONAL NEURAL NETWORKS</b><br/>
(sharing parameters, conv layers, pooling, CNN architectures)<br/>
<!--<a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_06.pdf" target="_blank">slides</a>-->
<hr/>
<b>PRACTICAL 1</b><br/>
Your first fully connected layer<br/>
gradient descent<br/>
stochastic gradient descent<!--<a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_Practical1_handout.pdf" target="_blank">slides</a>-->
</td>
<td class="dima">
19/10/18, 9am, QB.1.68- LECTURE 9<br/>
<b>RECURRENT NEURAL NETWORKS</b><br/>
(temporal dependencies, RNN, bi-directional RNNs, encoder-decoder, LSTM, gated RNN)
<hr/>19/10/18, 11am, QB.1.68- LECTURE 9<br/>
<b>RECURRENT NEURAL NETWORKS (CONT)</b><br/>
<br/>
<!-- <a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_07.pdf" target="_blank">slides</a>-->
</td>
<td class="labs">15/10/18, 12noon, QB.F101 - 2hr<br/> <b>Lab 2</b> - Your First Fully Connected Network</td>
</tr>
<tr> <td class="blank">4</td> <td class="dima">
22/10/18, 11am, QB.LT1.18 - <b>PRACTICAL 2</b><br/>
Error rate monitoring (training/validation/testing)<br/>
Batch-based training<br/>
Learning rate<br/>
Batch normalisation<br/>
Parameter intialisation<br/>
<!-- <a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_Practical2_handout.pdf" target="_blank">slides</a>--></td> <td class="tilo">22/10/18, 11am, QB.1.68- LECTURE 10<br/><b>THEORY SUMMARY AND QUESTIONS</b>
<hr/></td>
<td class="labs">22/10/18, 12noon, QB.F101 - 2hrs<br/><br/><b>Lab 3</b></td> </tr>
<tr> <td class="blank">5</td> <td class="dima">29/10/18, 11am, QB.F101 - <b>PRACTICAL 2 (cont)</b><br/> Batch-based training<br/>
Learning rate<br/>
Batch normalisation<br/>
Parameter intialisation
</td> <td class="blank">Invited Talk</td> <td class="labs">29/10/18, 12noon, QB.F101 - 2hrs<br/><br/><b>Lab 4</b></td></tr>
<tr> <td class="blank">6</td> <td class="dima">05/11/18, 11am, QB.F101 - <b>PRACTICAL 3</b><br/>Data augmentation<br/>
Debugging strategies<br/>
Hyperparameters (again)<br/>
<!-- <a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_Practical3_handout.pdf" target="_blank">slides</a>--></td> <td class="blank">Invited Talk</td> <td class="labs">05/11/18, 12noon, QB.F101 - 2hrs<br/><b>Lab 5</b></td> </tr>
<tr> <td class="blank">7</td> <td class="dima">12/11/18, 11am, QB.F101 - <b>PRACTICAL 4</b><br/> Baseline models<br/>
Adversal training<br/>
<a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_Practical4_handout.pdf" target="_blank">slides</a>
</td> <td class="blank">Invited Talk</td> <td class="labs">12/11/18, 12noon, QB.F101 - 2hrs<br/><b>Lab 6</b></td> </tr>
<tr> <td class="blank">8</td> <td class="blank" colspan="3">CS EXPLORE WEEK</td></tr>
<tr> <td class="blank">9</td> <td class="dima">26/11/18, 11am, QB.F101 - <b>CW Q/A</b>
</td> <td class="blank">Invited Talk</td> <td class="labs">26/11/18, 12noon, QB.F101 - 3hrs<br/><b>CW Lab</b></td> </tr>
<tr> <td class="blank">10</td> <td class="assess">03/12/18, 11am, QB.F101 - <b>Assessment Talks</b> (3hrs)</td> <td></td> <td></td> </tr>
<tr> <td class="blank">11</td> <td class="assess">10/12/18, 11am, QB.F101 - <b>Assessment Talks</b> (3hrs)</td> <td></td><td></td> </tr>
<tr> <td class="blank">12</td> <td class="assess">17/12/18, 11am, QB.F101 - <b>Assessment Talks</b> (3hrs)</td> <td></td> <td></td> </tr>
</table>
<hr/>
<h2>Assessment Details</h2>
<p>The student will undertake a challenge of replicating a
state-of-the-art performance on a publicly available dataset using one of
the deep architectures discussed on the unit. The unit has three
assessments:</p>
<ol>
<li> Lab Portfolio (individual, summative, 15%) [Wk10]</li>
<li> Talk Assignment (individual, summative, 25%) [Wk10-12]
<!--<a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_talk.pdf"><b>Talk Assignment</b> (individual, summative, 20%) [Wk10 - Wk12]</a> - Schedule for talks is now available at: <a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/COMSM0018%20-%20Applied%20Deep%20Learning%20Symposium.pdf">schedule (Wks10-12)</a>--></li>
<li> <!--<a href="https://www.ole.bris.ac.uk/bbcswebdav/courses/COMSM0018_2017/content/COMSM0018_Project_final.pdf">-->Group (up to 3) Project - assessed by a final report (summative, 60%) [Wk13] </li>
</ol>
<hr/>
<h2>Github</h2>
<p>All technical resources will be posted on the
<a href="https://github.com/COMSM0018-Applied-Deep-Learning" target="_blank">COMSM0018 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: COSC018263</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>