Deep Learning Study Group
General Information
Lecturer: | (various) |
Venue: | Thursday 15:30 – 17:00, Mediathek, B11 |
First Session: | November 12th, 2015 |
Sessions
- [November 12th, 2015]
Introduction and Planning. - [November 19th, 2015]
Reading: (Bengio, 2009) chapters 1-3. - [December 17th, 2015]
Reading: (Socher, 2015) Lecture Notes.
Upcoming sessions
- [tbd]
Distributed Representations of Words and Phrases and their Compositionality (Johannes) - [tbd]
Ad Hoc Monitoring of Vocabulary Shifts over Time (Michael) - [tbd]
Overview: Recursive und recurrent neural networks (Jonas) - [tbd]
GloVe: Global Vectors for Word Representation (Tim) - [tbd]
Improving Word Representations via Global Context and Multiple Word Prototypes (Martin) - [tbd]
Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank (Henning)
Literature
Video Lectures:
- Richard Socher, CS224d. Stanford University, 2015.
- Geoffrey Hinton, Neural Networks for Machine Learning. Coursera, 2012.
Textbooks and Chapters:
- Yoshua Bengio, Deep Learning. Book in preparation for MIT Press, 2015.
- Yoshua Bengio, Learning Deep Architectures for AI, Foundations and Trends in Machine Learning, Vol. 2, No. 1 (2009) 1–127.
- Hal Daumé III, A Course in Machine Learning, (preprint), 2015.
Surveys and Collections:
- Jürgen Schmidhuber, Deep Learning in Neural Networks: An Overview. Neural Networks, Volume 61, January 2015, Pages 85-117.
- Jürgen Schmidhuber, Literature Recommendations for New Researchers. Reddit.com, 2015.
- Yoshua Bengio et al., Deep Learning Tutorials. Deeplearning.net, 2015.
- Wojciech Samek, Hot Topics in Machine Learning: Deep Learning (Reading Material). TU Berlin, 2015.
- Christopher D. Manning, Last Words: Computational Linguistics and Deep Learning. COLI, MIT Press, 2015.
- Myungsub Choi, Jiwon Kim, Awesome Recurrent Neural Networks: A curated list of resources dedicated to recurrent neural networks, github.com.
Articles and Applications:
- Richard Socher et al., Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank. EMNLP'13.
- Giuseppe Attardi, DeepNL: a Deep Learning NLP pipeline. NAACL'15.
- Andrej Karpathy, The Unreasonable Effectiveness of Recurrent Neural Networks. 2015.
Software:
- Theano and pylearn2 (Python)
- Torch7 (Lua)
- Caffe (C++)
- Tensorflow (Python/C++)
- Keras (Python)