Homework solutions for Pattern Recognition and Machine Learning course at Tampere University. The data and the tasks themselves are provided. For more info about the course please visit the official web page.
Task 1 - A simple matplotlib plot
Task 2 - Fix an image with uneven brightening using a Gaussian
Task 3 - Estimate sinusoidal parameters
Task 1 - Implement a sinusoid detector
Task 2 - Train sklearn classifiers
Task 1 - Load a dataset of images and split them into training and testing
Task 2 - Train classifiers for the GTSRB task
Task 3 - Train ensemble methods with the GTSRB data
Task 0 (pen&paper) - create a keras neural network and compute the number of parameters
Task 1 - Define the neural network in keras
Task 2 - Compile and learn the network
Task 3 - Implement gradient descent for log-loss
Task 1 (pen&paper) - count the number of parameters of a given network
Task 2 (pen&paper) - count the number of parameters of a given network and compute the number of the multiplications on the last layer
Task 3 - MNIST with custom network
Task 4 - MNIST with pretrained network
Task 5 - Train the model
Task 1 (pen&paper) - Error rate confidence limits
Task 3 - download a high-dimensional ovarian cancer dataset and train a random forest classifier with 100 trees and plot a histogram of its feature importances
Task 4 - Apply the recursive feature elimination approach
Task 5 - Apply L1 penalized Logistic Regression for feature selection