Project 2 in FYS-STK4155 by Betina Høyer Wester, Mona Heggen and Henrik Gjestang. Course taken fall of 2018.
In this project, we studied 1D and 2D Ising model data with different machine learning techniques. More details may be found in the project report.
We use generated data sets and a data set from "A high-bias, low-variance introduction to Machine Learning for physicists" by Mehta et al. Since the latter is rather large, we randomly chose 10% of it in the readData script to use in the classification algorithms (saved as test_set.npy).
This script contains all the functions necessary to estimate the Ising energy from a generated set of spin variables. Ordinary Least Squares, Ridge Regression and the Lasso are used, and with the bootstrap algorithm, one can evaluate the fit using the estimate MSE, R2-score, bias and variance.
This file contains all the functions necessary for a binary classification of spin lattices using logistic regression and gradient descent methods.
Reads the data obtained from "A high-bias, low-variance introduction to Machine Learning for physicists". This data is used to train and evaluate the logistic regression classifier.
Tests the logistic regression method.
neuralNetwork is used for regression analysis. neuralNetwork_Classification is used for binary classification purposes. Both can have an arbitrary number of hidden layers, initialised with hiddenLayer.