Old Agendas

Agenda 8.4.15

Start with only 1 other object and learn interactions with it Try to implement some sort of "smart" exploration learning that chooses actions in order to improve predictions.

Agenda 23.4

Create Testsuits:

• Give target position (for block) and evaluate required number of training iterations and performance
• Active exploration (for fixed amount of time) before prediction task etc

I should focus on the prediction and planning of actions and their effects on objects, rather than try to learn concepts or something like that.

29.9:

• Thesis:
  • Malte likes the current concept
    • Inverse Model might be to detailed, consider moving at least part of that to realization
  • Figures appear decent
  • Problem specification might need to be removed if it becomes redundant from the introduction
• Evaluation:
  • Make statistics about different starting postions
    • Maybe even about different testing positions
  • Evaluate different settings
    • Different objects for testing than training
    • Different configurations
  • Concentrate on evaluating the entire concepts as a whole first
    • Evaluating supparts if it yields interesting results
    • If possible, compare model performances between different supparts