Skip to content

vaibhavnaagar/machineLearning-exercises

Repository files navigation

Machine Learning Exercises

ML (Stanford University)- Coursera

Exercise-1

  • Linear Regression: Cost function, gradient descent for one variable and multi variables, feature normalization
  • Normal equations

Exercise-2

  • Logistic Regression: Sigmoid function, cost function and gradient, learning parameters using fminunc, regulariztion.

Exercise-3

  • Multi-class CLassification: Regularized logistic regression with cost function and gradient, one-vs-all classification, one-vs-all prediction.
  • Neural Networks: Feedforward propagation and prediction.

Exercise-4

  • Neural Networks Learning (Hand-written digit recognition): Feedforward and regularized cost function, backpropagation including sigmoid gradient and random initialization, gradient checking, learning parameters using fmincg, visualizing the hidden layer.

Exercise-5

  • Regularized Linear Regression: Regularized linear regression cost function and gradient.
  • Bias v.s. Variance: Learning curves.
  • Polynomial regression: Learning polynomial regression, selecting λ using a cross validation set, computing test set error and plotting learning curves with randomly selected examples.

Exercise-6

  • Support Vector Machines: Linear classification, non-linear clasification using Gaussian Kernel.
  • Spam Classification: Preprocessing emails, extracting features from emails using vocabulary list, training SVM for Spam Classification and predicting emails as spam or non-spam.

Exercise-7

  • K-means Clustering: Finding closest centroids, computing centroid means, random initialization.
  • Image Compression with K-means: K-means on pixels.
  • Principal Component Analysis: Implementing PCA.
  • Dimensionality Reduction with PCA: Projecting the data onto the principal components, reconstructing an approximation of the data, run PCA on Face Image Dataset and reduces dimensions.

Exercise-8

  • Anomaly Detection: Estimating parameters for a Gaussian distribution, selecting the threshold 'ε' using cross-validation dataset.
  • Recommender Systems (Movie Rating): Collaborative filtering learning algorithm- cost function and gradient with regularization, learning movie recommendations using fmincg.