Deep Learners, Vectorize!
Our Virtual Fortnightly Machine Learning Meeting Notes for February 24, 2022
Hey Outliers,
We had an interesting lightning talk provided by Charles Davi. Charles talked about his software that he built called Black Tree. Black Tree is an automated machine learning algorithm where we can create and train a model and get the analysis of the model. This is done both effectively (high accuracy) and efficiently (very fast). It works efficiently because it uses Octave in the backend which is a lot more efficient for computation than Python. The free version is up so we can play with it and see it for ourselves.
We briefly discussed vectorization and how this feature is very crucial for machine learning, especially deep learning. Deep learning is just matrix multiplication and optimizing it will speed up the process. Vectorization is critical because it allows the ability to chop up the task of multiplying matrices into small chunks and run them simultaneously (or in parallel), saving a lot of time. Charles also wrote a research paper going in depth how he utilizes vectorization in the making of Black Tree.
Black Tree: https://blacktreeautoml.com/
Charles’ Research Paper: Vectorized Deep Learning
Community Updates
Kaggle
TPS - February 2022
Michael’s Notebook: EDA-LGBM-Optuna-CV-OOF
Book Study
Feature Engineering in Machine Learning
Changed the schedule to only Wednesday at 11AM-12PM in Discord #the-feynman
Having it everyday for half hour was a hard commitment
Working on discussing chapter and making a project based on what we learn in each chapter
Thank you folks for participating in our last town hall! There are still some admin work that needs to be done for the community and we are slowly making progress to get that completed. If you have any ideas or would like to create your own session, feel free to create it in our Discord! See you next time!
Main Photo by Alberto Bobbera on Unsplash