My slides and jupyter notebook for my random walk on physics talk on neural networks. To get them:
- You can download as zip from github
- Using the command line:
git clone https://github.com/lsgos/RWOP
To run and view the ipython notebook (ipynb) file requires jupyter: you can get this by
- using pip:
pip install jupyter - Installing anaconda, which will come with jupyter.
Once you have jupyter installed, open a terminal/command prompt, navigate to the directory where you downloaded this repository and run
jupyter notebook
to start the notebook interface.
In order to run all the code in the notebook itself, you're going to need to install some extra python libraries /programs:
Note that caffe is a huge pain to install, and I'm not totally sure it works at all on windows. There is a community macOS port though I think. If you are having trouble then there is some help here
In case you are just looking for my last slide and don't want to/can't manage to install all of this stuff, here's my 'further reading' list from the talk.
- Neural networks and deep learning: An online book that derives backpropagation from scratch and shows how to implement a neural net totally 'by hand' in numpy
- The unreasonable effectiveness of recurrent neural networks: A fantastic blog post on recurrent nets (which are very cool indeed) and some crazy stuff thats been done with them
- Atari A neural network plays atari!
- Grabber Farm: Robots learning to pick stuff up
- Colah's Blog: Whole series of posts about deep learning and neural nets, including some more mathematical treatment of data as manifolds.
- Tensorflow: An open-source neural net library, which has some great hands on tutorials.
- Caffe: Another one, more tailored for image recognition, which comes with pre-trained models. WARNING: CAFFE IS A NIGHTMARE TO INSTALL
- Torch: Yet another neural net library, with the quirk of being in lua.
- Scikit-Learn: Easy machine learning in python, has a nice set of tutorials too.
- Fork this talk on github: This entire presentation is a jupyter notebook, and I plan to make the source avaliable on my github page, so you can see and run all the code I used to generate this slideshow!