Machine Learning
for JS developers
Magic!
Ok magic but
let's look at source code...
No dependencies, no libraries.
Both game AND learning algorithm are just ~850 lines of code including comments.
I can read it!
Neural network
Evolution
Game
- Speed control
- Bird
- Pipe
- Start game
- Update frame data
- Draw frame
- Init game
- Genome
- Generate
- Breed & Mutate
- Generations
- Neuron & Layer
- Network
- Save & Load
- Compute
Yep.
You don't need math PhD and supercomputer to do some ML.
Learn
- Free without certificate
- Author: Andrew Ng, founder of Coursera
- Python is fine, but also translated to js
- Hardest math is matrix multiplication & derivatives
Use in node/browser
tensorflow.js (former deeplearn.js)
"A WebGL accelerated, browser based JavaScript library for training and deploying ML models."
"...we are actively working on creating node.js bindings to the TensorFlow C API"
Use third-party
Train or use pre-trained models in cloud.
Need image/voice recognition, text processing, translation, chatbots, etc – use cheap, cutting edge models from market leaders:
Play with it!
Thank you!
