BMAML

Re-implement of Bayesian Model-Agnostic Meta-Learning

Work in progress...

hyper-param tuning(nearly done)
Contrast with MAML
visualization of predictions
Experiemnts on other dataset
update README...

Each directory

For the full pipeline of continious adaptation, go to con.baseline. con.baseline contains how to run
con.nornn: Removes the rnn kernel from the full pipeline
con.baseline.stop: Stop the gradient of history particles in rnn kernel
con.Bmaml.eq: BMAML