I completed Andrew Ng's/Stanford University's machine learning course on Coursera, but instead of using the Matlab templates provided by the course, I implemented everything from scratch in Python.
Contents:
-
Ex1: Linear regression
With one variable and with multiple variables.
-
Ex2: Logistic regression
Including regularization.
-
Ex3: Multi-class classification and neural networks
Handwritten number recognition using two different methods: one-vs-all logistic regression, and forward propagating a pre-trained neural network.
-
Ex4: Neural networks learning
Build and train a neural network, including backpropagation, and use it for handwritten number recognition.
-
Ex5: Regularized linear regression and bias vs variance
Including learning curves and polynomial regression.
-
Ex6: Support Vector Machines
6.1: Demonstrate scikit-learn SVMs with linear and Gaussian kernels on some sample 2D datasets.
6.2: Build an email spam classifier using natural language processing and a scikit-learn SVM.
-
Ex7: K-means Clustering and Principal Component Analysis
7.1: Build a k-means clustering algorithm and use it for image compression.
7.2: Build a PCA algorithm and use it for image compression and for visualisation.
-
Ex8: Anomaly Detection and Recommender Systems
8.1: Anomaly detection using a multivariate Gaussian model. Precision, recall, F1 score.
8.2: Build a recommender system and use it to recommend movies.