In this code, we propose a different implementation fo the original paper https://arxiv.org/abs/1709.07124 under PyTorch. This architecture is constructed by unfolding the iterations of a sequential iterative soft-thresholding algorithm (ISTA) that solves the optimization problem for sparse nonnegative matrix factorization (NMF) of spectrograms. We name this network architecture deep recurrent NMF (DR-NMF)
I'm a passionate data engineer and data scientist with a particular focus on Distributed Systems and Recommender Systems. I worked on various project from regression task to LLM fine-tuning, passing by AWS deployment and testing. Additionally, I have a strong foundation in classical machine learning and possess the requisite mathematical prowess.
Currently, I'm delving into the realms of Data Streaming and web development.
Languages and Tools
Languages:
Python3
Scala
JS
Rust
Best frameworks and main libraries for Python3:
PySpark
Pytorch
Numpy
Pandas
Sklearn
Langchain
My tools for Data Manipulation & Visualisation:
Conda
Jupyter
Spark
MySQL
MongoDB
Postgres
SQLite
Plotly
Matpltlib
Environments, Testing, Other:
NodeJs
Git
Docker
Pytest
Postman
GitLab
OS:
Linux
Ubuntu
re-deep-recurrent-nmf-for-speech-separation-by-unfolding-iterative-thresholding's People