What my current understanding of ML allows me to understand, is that ML is a very wide and deep field, one that goes as far and maybe more than what you could think of. Therefore at my current knowledge level, it’s not the best idea to try and follow a streamlined area in this early. Since there exist a numerous things, eg: Computer vision, Natural Language Processing, Reinforcement Learning, Generative AI, Predictive Analysis, Robotics, etc; my current long-term goal is to expand my horizons. Play with a multitude of things to get myself climbing the peaks before deciding which one to build my cabin on.

My current experience lays on:

• Working on some fundamentals in Machine learning

  1. Implementing Restricted Boltzman Machine [Stacking into DBN] into pytorch
  2. Stacked Denoising Autoencoders (Both imported from a book I was reading, in which they were given in theano)
  3. Studying the basic architecture and techniques in CNNs

• Making models that make use of ML algorthms to do predictive analysis/ clustering or classification:

  1. Decision Trees
  2. Random Forest Regression
  3. XG Boost
  4. ADA Boost
  5. Linear and Logistic Regression
  6. K-Nearest Neighbors
  7. Naive Bayes
  8. K Means clustering
  9. Collaborative filtering

• Using transfer learning to train CNNs:

  1. EfficientNetB-Family of models
  2. Resnet-50 and other in Family

I’d like to think that I have amassed some idea of what this vast field has to offer and therefore, am looking into getting more accustomed to it in general with better ideas of the math behind the models/algorithms.

Math in ML - Mathematics in ML sits at the very core idea of the field’s identity. For me, ML itself is an embodiment of Mathematics finding a core use in computers (not like computers themselves are a living example of maths working things out). Therefore it’s an important step to always stay aware of the mathematical concepts and logic behind how and why things are working, so we don’t start treating our models and transformers as magic black boxes but smart functions whose parameters have been trained/set to do what they are doing.
Therefore one of my objectives is to further indulge in the mathematics. The book I’ve picked for now and have been using, is - “Machine Learning: A probabilistic Perspective” By Kevin P. Murphy.

Exposure in ML - I want to be exposed to more and more frontiers of ML even though I am working my way up from some fundamentals. One good method I’ve found is to throw myself at research papers and read them to try and see what’s cooking in the field.

That’s the plan for now and this post is to cement this idea somewhere more concrete and sturdy than the back of my sketchbook or my notes app.

Thanks for reading, Hope you have a nice day!