deep learning

Keras

• image_search: VGG-16 applied to Caltech101 dataset for nearest neighbor image retrieval
• keras_lenet: LeNet architecture for MNIST digit recognition
• keras_vae: Variational Auto-Encoder (VAE) trained on MNIST digits
• keras_dcgan: Deep Convolutional GAN (DC-GAN) trained on CIFAR10 dataset
• sentiment_kernel: LSTM neural network for sentence sentiment prediction
• lstm_language: LSTM language model for generating text trained on Shakespeare
• lstm_series: LSTM time series prediction applied to S&P500 data
• pretrained: pretrained VGG16, VGG19, ResNet50, InceptionV3 for image classification
• transfer_learning: ResNet50 fine tuned on Caltech101 dataset
• keras_seq2seq: seq2seq model for machine translation with bidirectional RNN encoder
• grad_cam: class activation map computed by weighing each channel by its average gradient
• style_transfer: neural style transfer with VGG19 by minimizing content and style loss with L-BFGS
• keras_mdn: Mixture Density Network (MDN) for learning parameters of a Gaussian Mixture Model
• keras_transformer: Transformer for text classification

References:
https://keras.io/

PyTorch

• lenet5_cifar10: LeNet5 CNN architecture for CIFAR10 object classification
• dan_sentiment: sentiment classifier based on averaging of pretrained word embeddings
• lstm_qsim: LSTM encoder for question similarity trained on stack exchange dataset
• gradients: computes gradient and weight norms for a simple MLP architecture with different optimizers
• normalizing_flows: a sequence of invertible density transformations for posterior approximation
• fine_tuning_BERT: fine tuning BERT for text regression task
• distilbert_imdb: fine tuning DistilBERT for IMDB sentiment classification using HuggingFace transformers
• zeroshot_vit: zero-shot image classification using HuggingFace transformers
• llama2_chat_langchain: integrating llama 2 chat with HuggingFace and LangChain
• peft_finetuning_llm: PEFT finetuning of FLAN-T5 model using LoRA adapters
• rag_openai_langchain: Retrieval Augmented Generation (RAG) using ChatOpenAI LLM and FAISS vectorstore

References: http://pytorch.org/

TensorFlow

• tf_classifier: DNN classifier for Iris dataset
• tf_regressor: DNN regressor for estimating Boston housing prices
• tf_mlp: Multi-Layer Perceptron with callbacks
• tf_cnn_mnist: CNN for MNIST digit classification
• tf_autoencoder: a two-layer encoder / decoder architecture applied to MNIST digits
• tf_word2vec: word2vec skip-gram model trained on the 20 newsgroups dataset
• tf_wide_and_deep: wide and deep classification architecture trained on census income dataset
• tf_gan_mlp: generative adversarial network based on two MLPs using MNIST digits
• tf_inception_v3: InceptionV3 architecture pre-trained on ILSVRC-2012-CLS image classification dataset
• tf_optimizers: a comparison of SGD, Momentum, RMSProp and Adam optimizers using a CNN trained on MNIST digits

References:
https://www.tensorflow.org/

Theano

• cnn_lenet: LeNet CNN architecture for MNIST digit classification
• mlp: Multi-Layer Perceptron
• logistic_sgd: Logistic Regression
• theano_linreg: Linear Regression

References: http://deeplearning.net/software/theano/

Python 2.7