Skip to content

Latest commit

 

History

History
53 lines (31 loc) · 1.08 KB

LOG.md

File metadata and controls

53 lines (31 loc) · 1.08 KB

Dec 02

13:57

Ran bayes_cbf.pendulum:learn_dynamics() Logs data/runs/learn_dynamics_v1.0.0-7-g718f8a5/events.out.tfevents.1606945888.dwarf.25367.0

Plots are okay, but need to plot from the logs. Also need to compare with other learning methods.

14:07

Saved logs were bad because fx and gx data were saved as same name; overridden.

Re-ran bayes_cbf.pendulum:learn_dynamics() Too much spread Logs: data/runs/learn_dynamics_v1.0.1-1-g97c87d8/events.out.tfevents.1606946860.dwarf.27863.0

Did not save Xtrain. Working around that.

15:12

Moved from plotting online to logging and then plotting.

Successful log data/runs/learn_dynamics_v1.0.2/events.out.tfevents.1606951767.dwarf.13002.0 Move to saved-runs/learn_dynamics_v1.0.2/events.out.tfevents.1606951767.dwarf.13002.0 Copied to overleaf with script.

Dec 03

11:43

  1. Compute ControlAffineRegressor with a different kernel.

Dec 07

11:45

  1. Decided not to further simplify Independent GPs

Dec 16

20:38

  1. Why is decoupled GP slower than MVGP

July 16th

12:28

  • Visualize in 3D
  • CBF? CLF? Yes for safety