search

tspooner / spaces

Set/space primitives for defining machine learning problems.
MIT License
9 stars 1 forks source link

Docs References #33

Open mohammedaugie13 opened 1 year ago

mohammedaugie13 commented 1 year ago

Hi, im kinda new in ML field can you give us the references or theorethical docs about this library. Thanks

tspooner commented 1 year ago

Hey, thanks for reaching out. This library is really very closely related to the spaces sub-module that is included in gymnasium; see here. From a practical perspective, that should give you some idea where to start looking.

In short, the main purpose is to specify the "spaces" on which problems are defined. For example, in reinforcement learning, one must specify what "valid" actions an agent can take in an environment. This could be, for instance, the set of positive integers...

With regards to "theoretical" resources, I would suggest getting started with some of the foundational texts in mathematical analysis and/or abstract algebra. One of my personal favourites is "Mathematical Analysis" by Tom Apostol which builds up some of the foundations you'll need. Mind you, I wouldn't claim that there's anything particularly deep in spaces, but understanding topics like "metric spaces" can be useful (contraction mappings are super important).

tspooner commented 1 year ago

As an extra thought, here are some of my favourite texts in the ML/stochastics space:

Enjoy!