DEV Community

Cover image for Laying the Foundation for Deep Learning
Sam Oz
Sam Oz

Posted on

Laying the Foundation for Deep Learning

This post is for those are lost along the way just like me.
Alt Text
Last week I have started a study marathon and the reason to do that I've realized I am a bit rusty about the foundations of the domain. I tried to find a study partner but since I couldn't find a suitable partner I will be using this platform as a journal to keep me in line and hopefully help couple of you along the way. About me; I am working as a data analyst at PwC(when I started writing this post) so I know statistics and machine learning as a practitioner. I am not coming from CS or Math background, I have studied Economics therefore I had some math knowledge kept me going for a while. I am also writing my thesis which is on natural language processing for that reason alone I must know the mathematical background of neural nets and modern architectures used in NLP such as transformers etc. to explain in academic lingo. Other than that NLP is rapidly growing and improving field and you should be able to read recent academic papers to stay up-to-date on state of the art methodologies. Additionally I should mention I know intermediate level Python and R programming so it won't be in my curriculum.
Last warning: As I mentioned above, I know most of the topics but need a refresher for foundations. So this is not a beginner-friendly walkthrough.

So enough of me, I will demonstrate my way of getting this done and I hope this helps people with similar background. Here is the path I am and will be walking on for a while:

1) Linear Algebra > 2) Multivariable Calculus > 3) Neural Networks > 4) Hyperparameter tuning, Regularization and Optimization > 5) Basic Architectures(like RNN,ConvNets, LSTM) > 6) NLP Basics > 7) Attention Models

Here are the resources for this steps:
1) LINEAR ALGEBRA
First part of the course is Multivariable Calculus
Math for Economists by Jason Kronewetter Ph.D.
http://ocw.uci.edu/courses/math_4_math_for_economists.html

Alternative:--You can also go for Gilbert Strang's Linear Algebra course on MIT open courseware.--
https://ocw.mit.edu/courses/mathematics/18-06-linear-algebra-spring-2010/

2) MULTIVARIABLE CALCULUS
Second part of the course is Multivariable Calculus
Math for Economists by Jason Kronewetter Ph.D.
http://ocw.uci.edu/courses/math_4_math_for_economists.html

Alternative:--MIT ocw is here for you again.-- https://ocw.mit.edu/courses/mathematics/18-02sc-multivariable-calculus-fall-2010/1.-vectors-and-matrices

3-4-5) DEEP LEARNING-OPTIMIZATION_ARCHITECTURES etc.
Andrew NG's legendary course will be enough for the theory.

https://www.coursera.org/specializations/deep-learning#courses

6-7) NATURAL LANGUAGE PROCESSING AND ATTENTION MODELS
Recently deeplearning.ai has released a great course for NLP practioners.
https://www.coursera.org/specializations/natural-language-processing?

Alternative:--Stanford ocw has released a course while ago but it can be math&programming heavy for some of you. So this is my alternative.
https://www.youtube.com/playlist?list=PLoROMvodv4rOhcuXMZkNm7j3fVwBBY42z

I will be sharing my updates about study in the next part.

___________________UPDATES______________________
I have finished Linear Algebra part in 8 days. I also took notes I am planning to share.

I started Multivariable Calculus part. Seems not so hard for now.

Top comments (0)