The problem of statistical learning. Training versus test (Vapnik-Chervonenkis dimension, training and generalization). Lineat model (linear, non-linear and logistic regression). What it is and how to detect and deal with overfitting. Machine learning principles: Ocam razor, sample bias and data snooping. Similarity-based methods (closest neighbor, radial basis functions, density estimation). Neural networks (MLP, training, approximation and regularization). Support vector machines. Aggregation methods. Selection of variables.
- Abu-Moustafa, Y.S., Magdon-Ismail, M. e Lin H-S. (2012). Learning from data. AML-Book.com.
- Devroye, L., Gÿofi, L., e Lugosi, G. (1996). A probabilistic theory of pattern recognition. Springer-Verlag.
- Mohri, M., Rostamizadeh, A., Talwalkar, A. (2012). Foundations of machine learning . MIT Press.
- Murphy, K. P. (2013). Machine learning: a probabilistic perspective. The MIT Press.
- Hastie, T., Tibishirani, R., & Friedman, J. (2002). The elements of statistical learning. Springer Series in Statistics.