Keras-based Machine Learning Models

This GitHub repository provides a comprehensive collection of Keras-based machine learning models a wide range of tasks.

Summary

Regression Tasks

• MNIST-from-Scratch: Multiclass classification from scratch on the MNIST handwritten digits dataset.

• MNIST-With-Custom-Metrics-and-Callbacks: Multiclass classification on the MNIST handwritten digits dataset with custom metrics and callbacks.

• MNIST-With-Custom-Training_Loop: Multiclass classification on the MNIST handwritten digits dataset with a custom training loop.

• MNIST-with-Custom-Loss: Multiclass classification on the MNIST handwritten digits dataset with a custom loss.

Sequence Data and Time Series Forecasting

• US_Diesel_Retail_Price forecasting: Time series forecasting on the US Diesel Retail Price dataset with fully connected and LSTM networks.

Natural Language Processing

• Deepl_lr_for_text: Covering NLP, including Text Processing (Normalization, Tokenization, Indexing, Embedding) Modelling (bag-of-words and sequence model approaches), and different model architectures (Dense, Bidirectional LSTM).

• The Transformer Architechture : Covering self-attention , multi-head attention ,the transformer encoder

• English to Spanish Translation with RNN and Transformers models : Covering sequence to sequence learning , RNN based encoder and decoder , Transformer decoder.