Deep Learning Models

A collection of various deep learning architectures, models, and tips for TensorFlow and PyTorch in Jupyter Notebooks.

Traditional Machine Learning

• Perceptron [TensorFlow 1] [PyTorch]
• Logistic Regression [TensorFlow 1] [PyTorch]
• Softmax Regression (Multinomial Logistic Regression) [TensorFlow 1] [PyTorch]

Multilayer Perceptrons

• Multilayer Perceptron [TensorFlow 1] [PyTorch]
• Multilayer Perceptron with Dropout [TensorFlow 1] [PyTorch]
• Multilayer Perceptron with Batch Normalization [TensorFlow 1] [PyTorch]
• Multilayer Perceptron with Backpropagation from Scratch [TensorFlow 1] [PyTorch]

Convolutional Neural Networks

Basic

• Convolutional Neural Network [TensorFlow 1] [PyTorch]
• Convolutional Neural Network with He Initialization [PyTorch]

Concepts

• Replacing Fully-Connnected by Equivalent Convolutional Layers [PyTorch]

Fully Convolutional

• Fully Convolutional Neural Network [PyTorch]

AlexNet

• AlexNet on CIFAR-10 [PyTorch]

VGG

• Convolutional Neural Network VGG-16 [TensorFlow 1] [PyTorch]
• VGG-16 Gender Classifier Trained on CelebA [PyTorch]
• Convolutional Neural Network VGG-19 [PyTorch]

ResNet

• ResNet and Residual Blocks [PyTorch]
• ResNet-18 Digit Classifier Trained on MNIST [PyTorch]
• ResNet-18 Gender Classifier Trained on CelebA [PyTorch]
• ResNet-34 Digit Classifier Trained on MNIST [PyTorch]
• ResNet-34 Gender Classifier Trained on CelebA [PyTorch]
• ResNet-50 Digit Classifier Trained on MNIST [PyTorch]
• ResNet-50 Gender Classifier Trained on CelebA [PyTorch]
• ResNet-101 Gender Classifier Trained on CelebA [PyTorch]
• ResNet-101 Trained on CIFAR-10 [PyTorch]
• ResNet-152 Gender Classifier Trained on CelebA [PyTorch]

Network in Network

• Network in Network CIFAR-10 Classifier [PyTorch]

Metric Learning

• Siamese Network with Multilayer Perceptrons [TensorFlow 1]

Autoencoders

Fully-connected Autoencoders

• Autoencoder [TensorFlow 1] [PyTorch]

Convolutional Autoencoders

• Convolutional Autoencoder with Deconvolutions / Transposed Convolutions[TensorFlow 1] [PyTorch]
• Convolutional Autoencoder with Deconvolutions (without pooling operations) [PyTorch]
• Convolutional Autoencoder with Nearest-neighbor Interpolation [TensorFlow 1] [PyTorch]
• Convolutional Autoencoder with Nearest-neighbor Interpolation -- Trained on CelebA [PyTorch]
• Convolutional Autoencoder with Nearest-neighbor Interpolation -- Trained on Quickdraw [PyTorch]

Variational Autoencoders

• Variational Autoencoder [PyTorch]
• Convolutional Variational Autoencoder [PyTorch]

Conditional Variational Autoencoders

• Conditional Variational Autoencoder (with labels in reconstruction loss) [PyTorch]
• Conditional Variational Autoencoder (without labels in reconstruction loss) [PyTorch]
• Convolutional Conditional Variational Autoencoder (with labels in reconstruction loss) [PyTorch]
• Convolutional Conditional Variational Autoencoder (without labels in reconstruction loss) [PyTorch]

Generative Adversarial Networks (GANs)

• Fully Connected GAN on MNIST [TensorFlow 1] [PyTorch]
• Convolutional GAN on MNIST [TensorFlow 1] [PyTorch]
• Convolutional GAN on MNIST with Label Smoothing [TensorFlow 1] [PyTorch]

Recurrent Neural Networks (RNNs)

Many-to-one: Sentiment Analysis / Classification

• A simple single-layer RNN (IMDB) [PyTorch]
• A simple single-layer RNN with packed sequences to ignore padding characters (IMDB) [PyTorch]
• RNN with LSTM cells (IMDB) [PyTorch]
• RNN with LSTM cells (IMDB) and pre-trained GloVe word vectors [PyTorch]
• RNN with LSTM cells and Own Dataset in CSV Format (IMDB) [PyTorch]
• RNN with GRU cells (IMDB) [PyTorch]
• Multilayer bi-directional RNN (IMDB) [PyTorch]

Many-to-Many / Sequence-to-Sequence

• A simple character RNN to generate new text (Charles Dickens) [PyTorch]

Ordinal Regression

• Ordinal Regression CNN -- CORAL w. ResNet34 on AFAD-Lite [PyTorch]
• Ordinal Regression CNN -- Niu et al. 2016 w. ResNet34 on AFAD-Lite [PyTorch]
• Ordinal Regression CNN -- Beckham and Pal 2016 w. ResNet34 on AFAD-Lite [PyTorch]

Tips and Tricks

• Cyclical Learning Rate [PyTorch]

PyTorch Workflows and Mechanics

Custom Datasets

• Using PyTorch Dataset Loading Utilities for Custom Datasets -- CSV files converted to HDF5 [PyTorch]
• Using PyTorch Dataset Loading Utilities for Custom Datasets -- Face Images from CelebA [PyTorch]
• Using PyTorch Dataset Loading Utilities for Custom Datasets -- Drawings from Quickdraw [PyTorch]
• Using PyTorch Dataset Loading Utilities for Custom Datasets -- Drawings from the Street View House Number (SVHN) Dataset [PyTorch]
• Using PyTorch Dataset Loading Utilities for Custom Datasets -- Asian Face Dataset (AFAD) [PyTorch]
• Using PyTorch Dataset Loading Utilities for Custom Datasets -- Dating Historical Color Images [PyTorch]

Training and Preprocessing

• Generating Validation Set Splits [PyTorch]]
• Dataloading with Pinned Memory [PyTorch]
• Standardizing Images [PyTorch]
• Image Transformation Examples [PyTorch]
• Char-RNN with Own Text File [PyTorch]
• Sentiment Classification RNN with Own CSV File [PyTorch]

Parallel Computing

• Using Multiple GPUs with DataParallel -- VGG-16 Gender Classifier on CelebA [PyTorch]

Other

• Sequential API and hooks [PyTorch]
• Weight Sharing Within a Layer [PyTorch]
• Plotting Live Training Performance in Jupyter Notebooks with just Matplotlib [PyTorch]

Autograd

• Getting Gradients of an Intermediate Variable in PyTorch [PyTorch]

TensorFlow Workflows and Mechanics

Custom Datasets

• Chunking an Image Dataset for Minibatch Training using NumPy NPZ Archives [TensorFlow 1]
• Storing an Image Dataset for Minibatch Training using HDF5 [TensorFlow 1]
• Using Input Pipelines to Read Data from TFRecords Files [TensorFlow 1]
• Using Queue Runners to Feed Images Directly from Disk [TensorFlow 1]
• Using TensorFlow's Dataset API [TensorFlow 1]

Training and Preprocessing

• Saving and Loading Trained Models -- from TensorFlow Checkpoint Files and NumPy NPZ Archives [TensorFlow 1]