- Models built automatically for 104 brains are available already
- Consider rebuilding these with (what is currently) the improved algorithms
- Arbitrate the choice of modes so average evaluation for a range of "number of modes", e.g. 5-20
- Use shuffle distance with radius ~2.5, as before. Base this choice on the experiments in Part 1 so that the choice is justifiable.
- Needs intensive work on porting formats and writing code
- Can register the 104 brains using MI, NMI, CR, etc.
- Evaluate in a way that is consistent with the above evaluations
- Perform comparisons and draw some conclusions