Deep learning models for image data [in dl_vision/]

Transfer learning and interpretation on the Cats-vs-Dogs dataset (cats_vs_dogs_models/)

Models

• Vanilla CNN (baseline) - 71% acc cats_vs_dots_cnn.ipynb
• Vanilla CNN with data augmentation and dropout - 85% acc cats_vs_dots_cnn.ipynb
• Fine-tuned VGG19 model with data augmentation and dropout - 93% acc cats_vs_dots_VGG19.ipynb
• Fine-tuned ResNet50 model with data augmentation and dropout - 96% acc cats_vs_dots_ResNet50.ipynb

Model interpretation

• Visualizing layer activations on a test image for fine-tuned VGG19 and ResNet50 cats_vs_dots_layer_activations.ipynb
• Filter visualization using gradient-ascent for fine-tuned ResNet50 cats_vs_dots_grad_ascent.ipynb
• Grad-CAM visualization for fine-tuned ResNet50 cats_vs_dots_grad_cam.ipynb

Deep Dream (deep_dream/)

• Deep Dream

Classification and generation on the Fashion-MNIST dataset (fashion_mnist_models/)

Classification models

• Vanilla CNN - 91% acc.

Latent-space generative models

• Convolutional variational autoencoder for image generation

Machine learning algorithms for tabular data [in ml_tab/]

Classification (supervised/classification_basic.ipynb, supervised/classification_adv.ipynb)

• k-Nearest Neighbours classification
• Logistic Regression
• Logistic Regression using polynomial features
• SVM with linear kernel, binary classification, multi-class classification
• Decision Tree
• Gaussian Naive Bayes
• Random Forest

Regression (supervised/regression.ipynb)

• k-Nearest Neighbours regression
• Linear Regression
• Ridge Regression (Regularization)
• Ridge Regression with feature scaling
• Polynomial Regression

Clustering (unsupervised/clustering.ipynb)

• k-Means Clustering
• DBSCAN
• Spectral Clustering
• Markov Clustering (markov_clustering Python module)