Programming languages used: C, R Each folder contains asssignment solutions along with a detailed report on the experiments carried out (in Spanish)
Assignment 0: Generate spiral-like datasets
Assignment 1: Decision Trees
- Effect of training dataset size
- Overfitting and prunning
- Resistance to noise
- Effect of dimensionality
- XOR Problem
Assignment 2: Neural Networks
- Use ANN to the two-ellipses dataset. Separation between training, validation and test datasets.
- Effect of momentum and learning rate
- Effect of the number of epochs
- Use ANN to the two-spirals dataset.
- Effect of the number of neurons
- Effect of the number of samples in traing set
- Implementation of 'weight decay' for ANNs ( code in C )
- Effect of dimensionality
Assignment 3: Naive Bayes
- Implementation of Naive Bayes classifier ( code in C )
- Effect of dimensionality
- Performance in non-gaussian data
- Implementation of Naive bayes clf using histograms
Assignment 4: KNN
- Implementation of k-Nearest Neighbors ( code in C )
- Effect of dimensionality on KNN
- Effect of k
- Implementation of k-NN using voting radius ( code in C )
Final project: Application of Decision Trees, Bayesian learning and Support Vector Machines to Heladas dataset.
- Use k-fold CV to estimate performance of each classifier
- Experiment with different kernels for SVM
- t-tests