TRIPODS Data Science Boot Camp:
Building, analyzing, and interpreting machine learning models

Matthew Stone, John Blackmore and Dhruv Metha Ramesh

Summer 2022, June 21-23, 3pm-5pm

DIMACS seminar room, CoRE Building, 4th floor, Room 431

The DATA-INSPIRE TRIPODS Institute, in collaboration with the DIMACS REU program, organizes a Data Science boot camp for training of students in data science concepts.

In this series of hands-on sessions, we will explore techniques for using machine learning methods in data analysis.

– The first session will illustrate some general techniques for preparing data, building models, and assessing performance, with a specific case study on processing text sequences.

– The second session will look at visualization and interaction design patterns that you can use to explore models, and show how that exploration can assist in debugging models and pinpointing errors that suggest directions to improve the performance of machine learning models.

– The third session will look at how to interpret the output of models, and will particularly focus on the problem of calibrating probabilistic predictions from learned models so model output accurately reflects the confidence in a decision.

Each session will involve working with python notebooks in the google colab environment (a web based framework that requires no setup on individual machines), and will include an initial presentation of a worked-out case study and continue with a chance for students to work together to address related problems on their own. No background in machine learning, statistics or data science is expected. Some familiarity with python will probably be helpful, but python is very approachable and forgiving.