Deep Learning RTP Meetup - Getting Started With Deep Learning
Below are my notes from last night’s meet up with Deep Learning RTP. The presentation was an overview of Fast.ai - Practical Deep Learning for Coders MOOC. I remember the organizer pitching the course last year at a Pydata meetup. The group seemed to learn a lot in a good way in a short amount of time.
The course notes that you will probably spend 10 hours a week for 7 weeks which includes a total of 20 hours of video. I am concerned that I don’t have time to keep up with the group, but if I focus other goals (write, Python daily, learn something every week) around the course it might work out in the long run.
Last night’s presenter, Danny Siegel, had a good outlook “Try it and see if you like it.” Which is what I think I will do.
Fast.ai in 5 minutes
- About a year old
- People with stats and math backgrounds go faster
Fast.ai - Practical Deep Learning for Coders
- Videos, Notes, Forums
What is Deep Learning?
- Self Driving Cars
- Anything with Numbers
- Fully Connected Layers
- Activation Functions
- Stochastic Gradient Descent
- GPU (AWS), Cresle, Paperspace
- Kaggle data
Group is collaborative, community focused.
- Come as you are
- New People every week
- Everyone is student and teacher
- Supportive Environment
- infinitely flexible Functions
- Functions that map inputs to outputs
- Biggest promise is for unsolved problems
Services with GPUs
- 1 hour free
- Start Jupyter
- Use the terminal to clone the fast.ai GitHub repo
- Course directory is ready to go
- Choose Linux
- Couple hours to get approved, tell them doing the fast.ai Course
- Linear algebra focused
- distill pub for more basic fundamentals
- magenta TensorFlow sketch