This repository contains several deep learning projects, each focusing on different applications and models within the field of deep learning.
- Project 1: Clustering with SOM, Classification with SLFN
- Project 2: Recurrent Neural Networks in the World of Stocks
- Project 3: Blind Source Separation Using Variational Autoencoders
- Project 4: Persian News Classification Using LSTM+Attention
- Task:
- Part I: Use Self-Organizing Maps (SOM) for clustering countries based on coronavirus cases.
- Part II: Implement a Single Layer Feed-forward Network (SLFN) for classification tasks.
- Model:
- SOM for unsupervised clustering.
- SLFN for classification.
- Dataset:
- Coronavirus cases data for clustering.
- A provided dataset for classification.
- Task:
- Use simple RNNs (Elman or Jordan networks) to predict the Tehran Stock Exchange Index based on historical price data.
- Model:
- Simple RNNs (Elman or Jordan networks).
- Dataset:
- Four years of historical data for the Tehran Stock Exchange Index.
- Task:
- Part I: Use Variational Autoencoders (VAEs) to separate mixed MNIST and Fashion MNIST images.
- Part II: Use VAEs to separate vocal and background music components from audio recordings.
- Model:
- Variational Autoencoders (VAEs).
- Dataset:
- MNIST and Fashion MNIST for image separation.
- IRMAS dataset for music separation.
- Task:
- Implement an LSTM model with an attention mechanism to classify a Persian news dataset into different categories.
- Model:
- LSTM with attention mechanism.
- Dataset:
- Persian news dataset containing titles, summaries, and content of news articles across six categories.