This is my implementation of the coding homework in CS 189.
-
Used SVM to classify the MNIST dataset, spam dataset and CIFAR-10 dataset.
-
Finished hyperparameter tuning and used k-fold cross-validation to validate the model.
- Plotted the isocontours of several functions.
- Computed eigenvectors of the Gaussian Covariance Matrix
- Utilized LDA and QDA to classify the digits and spam dataset
- Implemented logistic regression with L2 regularization with gradient decent and stochastic gradient descent.
- Varied the learning rate of stochastic gradient descent.
- Designed a decision tree and random forest.
- Utilized the tree model to classify the spam and Titanic dataset.
- Visualized the MDS189 dataset.
- Implemented a muti-layers fully connected network.
- Constructed a CNN to complete the classification task.
- Used Singular Value Decomposition (SVD) to get the rank-k approximation.
- Ran the approach on two different images with different ranks.