PA196: Pattern Recognition
Exercises: Sept. 24, 2014
Dr. Vlad Popovici
popovici@iba.muni.cz
Institute of Biostatistics and Analyses
Masaryk University, Brno
Vlad PA196: Pattern Recognition
1. Perceptron
1 simple implementation of perceptron with discussion
2 discussion on the sklearn.linear_model.Perceptron
To do:
select one dataset (e.g. iris) and transform it to a binary
problem
check the help for the Perceptron class
use fit method to build a model
use the predict method to predict the labels of a data set
shuffle (numpy.shuffle()) the data and repeat the training:
is there any difference?
randomly partition (have a look at random.choice())
partition your data into a training set and a validation set
train a new model and test it on the validation set
Vlad PA196: Pattern Recognition
2. Fisher Discriminant
FDA is LDA for 2 classes
check sklearn.lda.LDA class
go through the code http://scikit-learn.org/stable/
_downloads/plot_lda_qda.py
repeat the training/testing steps from perceptron, with the FDA
classifier
Vlad PA196: Pattern Recognition