I am looking for a project to do over the summer to get acquainted with some tools used in the field of machine learning. Ive heard a lot of things about Tensorflow, would this be a good starting point? Google is not giving me much to work with, what are interesting things to do with it?
Bump. Projects using libraries other than TensorFlow are welcome as well, interested in anything machine learning actually.
did someone say tensor?
>>8914074
https://www.youtube.com/watch?v=vq2nnJ4g6N0
This video does well to introduce the applications.
>>8914074
Also check out sklearn if you're a python kinda anon
To get a good introduction, I would just find some tutorials on the methods/languages you're interested in and implement them.
Just grab some data from the UCI Machine Learning Repo. and try to predict/classify some stuff.
Implement an ANN, Support Vector Machine, K-Nearest Neighbors Classifier, K-Means Squared, CNN, RNN, whatever you're interested in then apply it to whatever dataset you want.
I've found some pretty good tutorials on YouTube for implementing an ANN and you only really need to do it once and apply it to any dataset (if you do it right).
Tools in particular, TF is a good start and like already mentioned sk-learn, for python. There are some C++ libraries for Genetic Algorithms and Neural Nets on GitHub as well. I'm sure you could find some libraries for any language. But to really get into ML I would recommend at least trying to implement some of these things from scratch.
UCI Repo: https://archive.ics.uci.edu/ml/
>>8914511
Going to watch this when I find the time, thanks!
>>8914528
Yeah I think I will be dealing Python mostly as its what I know best. May try my hand at R as well since it seems widely used, never hurts to pick up some extra marketable skills.
>>8914559
This ticks a lot of boxes, thanks. Im starting to form an idea of what approach (and to what specifically) I want to take, and the concepts you mentioned seem to connect with what I will be needing.
On a sidenote, Ive heard on more than one occasion that machine learning is tied to differential geometry. Can anyone tell me more about this? Results on Google are obtuse and not in-depth. In particular, as I am a maths students currently studying Lie groups, do Lie groups find any sort of application in machine learning?
>>8914074
You won't actually learn anything beyond learning to use software if you go this route. Barring some fundamental leaps in mathematics, Deep Learning is a scientific/computational dead end.
>>8914826
>On a sidenote, Ive heard on more than one occasion that machine learning is tied to differential geometry. Can anyone tell me more about this? Results on Google are obtuse and not in-depth. In particular, as I am a maths students currently studying Lie groups, do Lie groups find any sort of application in machine learning?
Manifold learning. It was the hot shit before Neural networks came around for the 3rd time and destroyed all progress in the field of machine learning.
https://www.cs.nyu.edu/~roweis/lle/
http://yeolab.weebly.com/uploads/2/5/5/0/25509700/belkin_laplacian_2003.pdf
http://wearables.cc.gatech.edu/paper_of_week/isomap.pdf
>>8914892
Got it, thanks!
Bump
I'm also a programmer aspiring to work on AI in the future, and I'm testing my ideas by trying to make robots in Garry's Mod that act like humans. For me it's very effective, I'd recommend it.