30 days of Deep Learning Fundamentals
What’s this about?
After years of going into lectures and yawning in classes, I finally decided to start a hands-on project on deep learning. Also I figured, why not document my progress along the way? I have bad memory anyways, and hopefully it will do some greater good to help a lost soul find its way home in the deep swamp of machine learning.
What’s your plan?
I will choose a range of hot topics like CV and NLP and find a couple of well-defined datasets for each of the topics. I will then work on these datasets and implement a number of DL models, tune the models, and analyze the results. The models should cover both simple architectures like MLP and the state-of-the-arts. Without doubt I will face lots of problems and questions during the process. I’ll try to dig into details and find answers and solutions to them and document what I found. It should be of a Q&A style but if some topics go beyond the Q&A scope, I’ll find a more appropriate writing style for those topics.
What’s the purpose?
Although ML in general is more about researching for better methodologies and architectures, I found it hard to believe that doing it only on paper is sufficient. By implementing exisiting models and tuning them to realize the results others have obtained, I may be able to have a deeper understanding in the related areas.
Do you really think you can finish it in 30 days?
The “30 day” is more of a definition of learning scope than a definition of time frame. It means I will focus only on the fundamental concepts and models, and make it possible to finish in 30 days if one decides to do a similar project, work full-time and do not procrastinate like me. I’m in my last year of uni so I can’t work full-time on this project just yet, but I will try my best to ensure that the memos have a better-than-average quality.
The Juicy Content
Computer Vision
Day 1 - Day 2
Multilayer Perceptron on MNIST