Skip to content

Latest commit

 

History

History
25 lines (12 loc) · 784 Bytes

TODO.md

File metadata and controls

25 lines (12 loc) · 784 Bytes
Raw

TODO

  1. Train segment model to sematically segment frames
  • notebook works, organize segmentation methods into class

  1. Create behavior tree for apply user generated constraints

  2. Create progressive perceptive field and occupancy detection for agent

  3. Create tokens and transformer for actions

  4. Implement DQN for filtering action tokens

  5. Create method for discovering motifs

  • get vertical position of mario on screen, feed into matrixprofile method for action classification (STAMP, STOMP, SCRIMP+, ???), see if forward/backward move commands can be added as another dimension

  1. Build system for training new DNNs on sequence clusters

  2. Build system for generating new tokens and retraining transformer

  3. Evaluation method for deterritorializing bad tokens