Decisions, data and machine learning

DATES: 14 - 18 October 2019 (5 sessions)
TIME: 09:00 - 11:00 (a total of 10 hours)
LOCATION: BCAM Seminar room

ABSTRACT:
Data serves to improve decision making. Data has been shown to be an extremely useful resource in many fields including communications, weather forecasting, and economics. Modern data processing techniques developed under paradigms such as machine learning, data science, and artificial intelligence are enabling many critical applications. This course provides an introduction to data-driven decision problems from classical decision/game theory pioneered in the 40s by von Neumann to its application in modern machine learning problems.

PROGRAMME:
1. Decision problems
1.a. Consequences and preferences
1.b. Actions, states, and loss functions
1.c. Entropy and divergence (regret)

2. Data-aided decision problems and supervised classification
2.a. Value of data
2.b. Minimax actions
2.c. Maximum entropy, exponential families, and logistic regression

PREREQUISITES:
Basic knowledge of probability calculus and linear algebra.

REFERENCES:
[1] D. V. Lindley, Making decisions. John Wiley and Sons, London, 1985.
[2] Peter D. Grünwald and A. Philip Dawid. "Game theory, maximum entropy, minimum discrepancy and robust Bayesian decision theory." Annals of Statistics, 32(4): 1367-1433, 2004.
[3] M. Mohri, A. Rostamizadeh, and A. Talwalkar, Foundations of Machine Learning. MIT Press, Second Edition, 2018.

*Registration is free, but mandatory before October 10th.
To sign-up go to https://forms.gle/huniytjSHQt7Jmk56 and fill the registration form.

Student grants are available. Please, let us know if you need support for travel and accommodation expenses in the previous form before September 15th.

• BCAM_UPV_Course_2019_10_14