stat212b
Topics Course on Deep Learning for Spring 2016
UC Berkeley, Statistics Department
##Syllabus
1st part: Convolutional Neural Networks
- Invariance, stability.
- Variability models (deformation model, stochastic model).
- Scattering
- Extensions
- Group Formalism
- Supervised Learning: classification.
- Properties of CNN representations: invertibility, stability, invariance.
- covariance/invariance: capsules and related models.
- Connections with other models: dictionary learning, LISTA, Random Forests.
- Other tasks: localization, regression.
- Embeddings (DrLim), inverse problems
- Extensions to non-euclidean domains.
- Dynamical systems: RNNs and optimal control.
- Guest Lecture: Wojciech Zaremba (OpenAI)
2nd part: Deep Unsupervised Learning
- Autoencoders (standard, denoising, contractive, etc.)
- Variational Autoencoders
- Adversarial Generative Networks
- Maximum Entropy Distributions
- Open Problems
- Guest Lecture: Ian Goodfellow (Google)
3rd part: Miscellaneous Topics
- Non-convex optimization theory for deep networks
- Stochastic Optimization
- Attention and Memory Models
- Guest Lecture: Yann Dauphin (Facebook AI Research)
Schedule
Lec1 Jan 19: Intro and Logistics
Lec2 Jan 21: Representations for Recognition : stability, variability. Kernel approaches / Feature extraction. Properties.
Lec3 Jan 26: Convnets / Scattering
Lec4 Jan 28: Scattering Properties
Lec5 Feb 2: Further Scattering
Lec6 Feb 4: Supervised Learning: classfication. Properties of learnt representations
Lec7 Feb 9: Properties of learnt representations. Covariance and Invariance. Estimation properties: above generalization error. Redundancy in parameter space.
Lec8 Feb 11: Connections with other models (DL, Lista, Random Forests, CART)
Lec9 Feb 16: Representations of stationary processes. Properties.
Lec10 Feb 18: Other high level tasks: localization, regression, embedding, inverse problems.
Lec11 Feb 23: Extensions to non-Euclidean domain. Sequential Data RNNs.
Lec12 Feb 25: Guest Lecture ( W. Zaremba, OpenAI )
Lec13 Mar 1: Unsupervised Learning: autoencoders. Density estimation. Parzen estimators. Curse of dimensionality
Lec14 Mar 3: Variational Autoencoders
Lec15 Mar 8: Adversarial Generative Networks
Lec16 Mar 10: Maximum Entropy Distributions
Lec17 Mar 29: Self-supervised models (analogies, video prediction, text, word2vec).
Lec18 Mar 31: Guest Lecture ( I. Goodfellow, Google Brain )
Lec19 Apr 5: Non-convex Optimization: parameter redundancy, spin-glass, optimiality certificates. stability
Lec20 Apr 7: Tensor Decompositions
Lec20 Apr 12: Stochastic Optimization, Batch Normalization, Dropout
Lec21 Apr 14: Reasoning, Attention and Memory: New trends of the field and challenges. limits of sequential representations (need for attention and memory). modern enhancements (NTM, Memnets, Stack/RNNs, etc.)
Lec22 Apr 19: Guest Lecture (Y. Dauphin, Facebook AI Research)
Lec23-25 Oral Presentations