ML+ web dev (learning):
https://m.youtube.com/playlist?list=PLOU2XLYxmsILr3HQpqjLAUkIPa5EaZiui
Narrow AI : trained to do specific thing like or even better than humans.
Machine Learning : An approach to AI where system learns from patterns.
Deep Learning : A technique to implement Machine Learning.
Reinforcement Learning : try to take actions to maximize its reward.
Transfer Learning : retrain existing models with new data.
How to train ML model :
- Features and Attributes (color shape size weight position etc.)
- Visualizing features
- Choose an algorithm
TensoFlow.js :
Tensorflow.js is high level Layers API (like Keras(high level layers API for python)) which was build after deeplearn.js which was a low level mathematical Ops API which required more knowledge of ML and Math to run a model on browser.
Both Python and JavaScript code for Tensorflow.js are simply build on top of C++ core with the help of C/C++ bindings.
Models in Tensorflow.js can run on both client and server side. Client side our hardware have many options and execution time of ML model is based on hardware of client. While on server side hardware is fix and mostly of better quality and better scalability options
Client Side hardware options :