I am currently finishing up my undergraduate degree in Aerospace Engineering and am very interested in the application of ML for Physics informed and Scientific purposes such as Digital Twin efforts and Design optimization. I have taken quite a few background online courses and watched many lecture series on teh basics of ML as well as currently working through the Deep Learning course. I plan to continue onto IBMs courses on Pytorch, TensorFlow, and Keras but I don’t know where to go from there. I am also currently involved in an internship that will likely transition to industry applications of this but I want to use my spare free time to get better at the application of ML. Do you have any recommendations on where I can even start. It seems that most recommendations are to start trying to replicate Research Papers but that seems to be too far outside of my wheel house to start and I want to work my way to that point. Thanks!
You can get additional hands-on experience by going to a site like Kaggle, which has tutorials and example datasets you can work on.
Essentially, get some datasets (there are many available online for free), and create models for those.
I guess I should add that while I am aware of datasets, I don’t necessarily know how to start with them. Does kaggle provide walk throughs for these as well that improves my general ability to start on a project without potential handholding from tutorials
Pretty much every site that provides tutorials will give you a dataset that has already been pre-processed to have correct labels and consistent formatting.
A tutorial will typically teach a specific ML method.
You’ll need experience with each type of method (or model) that you might want to use for a specific task.
- The key to making good models is experience.
- The key to experience is practice (and asking questions when you are stuck).