Institute for Applied Computational Science

Sheila Coveney, Program Manager

617-384-9091

iacs-info@seas.harvard.edu

Data Science in Python (Day Two)

Tuesday, January 10, 2017

8:30 AM - 11:30 AM

This is day two of a five-day, three hours-per-day workshop that will take you from being a person with some idea of how to program to a person with some idea of how to do data science. Day two will cover the following topics: Exploratory analysis and visualization; the basics of machine learning.

Pre-requisites:

Attendees must have programmed in some programming language; being math savvy will help but is not necessary.

Participants must bring a laptop with Anaconda Python Distribution installed: https://www.continuum.io/downloads. We will use Python 2.7 in all sessions.

**Overview of the entire week:**

We'll work through learning those parts of Python needed to do data science, starting with numerical python; we'll then move on to exploratory data analysis and visualization; from there we'll tackle training some machine learning models, both regression (the prediction of continuous outcomes) and classification (the prediction of labels), including concepts such as feature selection, cross-validation, and regularization, and (time permitting) including the use of ensembles.

Finally, you’ll learn how to train these models when the data sizes are two large for one machine, and how to reduce the amount of computational time required to train these models.

Topics covered throughout the week include:

Day 1: Monday, January 9

Intro to Python, Numpy, Matplotlib, and Bokeh.

Day 2: Tuesday, January 10

Exploratory analysis and vizualization; the basics of machine learning.

Day 3: Wednesday, January 11

Learning a model (complexity, regularization, cross-validation); Regression.

Day 4: Thursday, January 12

Classification and Model comparison.

Day 5: Friday, January 13

Large scale machine learning with joblib, dask, and ipython parallel. If time permits, Ensembles.

*Please note that you must register for each workshop separately.*