May 5 & 12, 2017: Machine Learning (James Sharpnack)
May 5: Part I - Lessons from Binary Classification: Overfitting and surrogate losses (James Sharpnack) Binary classification is the canonical machine learning task and its study has a rich history. Much of the main principles of machine learning have been discovered in this context. These principles continue to guide the broader development of machine learning, and they motivate methods such as cross-validation, kernel support vector machines, logistic regression, and neural networks. We will highlight the resultant methods and will accompany this with data driven examples. To best prepare for the workshop, bring a laptop pre-loaded with jupyter, python, numpy and scikit-learn (they can be installed through anaconda, https://anaconda.org/anaconda).
May 12: Part II - Lessons from Binary Classification: Non-parametric methods and the kernel trick (James Sharpnack) In this second session, we will continue to explore machine learning with data driven examples.