There will be some introductory meetings, tutorials and paper readings. For paper reading days, we will have two papers everyday and discussion. Each student will present two papers, one basic introductory (relatively easy) and one paper involving machine learning as well as differential privacy. Please select your paper from list in the material section. The list in the material section is for suggestion and is not an exhaustive list. So you can also suggest a different paper (which other should agree) to be included in the list that you may like to present or you may want someone else to present.
- Introduction of course members
- Free format but would like to hear about Name, background, objective or expectation from the course
- What does the title mean
- What you can expect to learn at the end of the course
- How we should organize the course
- We would like to hear about your thought on the plan
- Should we cover some initial materials or tutorials e.g. probability or machine learning for example classification, regression
- How we can have an active discussion group
Machine learning basics
* What is ML?
* data, model and prediction
* Machine Learning tasks and domains, examples:
* Supervised learning
* Unsupervised learning
* Dimensionality reduction
* Linear regression, regularization/penalization
* Classification: logistic regression
* K-means clustering
* Principal components analysis
* Generalization (prediction, generalizability of inferences, training set error, test set error), cross-validation, parameter tuning, model selection
* Probabilistic modelling
* Basic probability distributions (random variables, parameters)?
* Why probability? Quantification of uncertainty (limits of data, limits of knowledge).
* Maximum likelihood learning
* Bayesian learning
- Different privacy mechanisms
- Secret sharing
- Definition, Sensitivity
- Laplace mechanism, Gaussian mechanism
- Properties - composition, plausible deniability, composition
- Examples - counting, sum, mean, Naive Bayes
- Exponential mechanism
- Local sensitivity,
- Redundant queries
- Robust De-anonymization of Large Sparse Datasets. Arvind Narayanan and Vitaly Shmatikov. IEEE Symposium on Security and Privacy, 2008.
- Thanh Bui
- Introduction to multi-armed bandit problems
- Differential privacy and multi-armed bandit problems
- Thanh Bui
- Project presentations