Build: #176 was successful Changes by Marcus Lewis <>

Build result summary


10 minutes
bamboo-agent-3 (39)
43e3ef94b54f16c3aa77b55bd2e5d9822720cf3d 43e3ef94b54f16c3aa77b55bd2e5d9822720cf3d
abd1ec2da543f9e84029f8d003154829738a40c0 abd1ec2da543f9e84029f8d003154829738a40c0
Successful since
#167 ()

Code commits

Author Commit Message Commit date
Marcus Lewis <> Marcus Lewis <> abd1ec2da543f9e84029f8d003154829738a40c0 abd1ec2da543f9e84029f8d003154829738a40c0 Merge pull request #870 from mrcslws/gaussian-bumps
Add model that more closely mimics rat entorhinal modules
Marcus Lewis <> Marcus Lewis <> 81df705d64a4359544a53eb1aade146c938e91ed 81df705d64a4359544a53eb1aade146c938e91ed Experiment parameters: s/cellsPerAxis/locationModuleWidth
Marcus Lewis <> Marcus Lewis <> 7db621ccadd90a15076820cd936ee1450ddb76ce 7db621ccadd90a15076820cd936ee1450ddb76ce Add model that more closely mimics rat entorhinal modules.
Here's what distinguishes this model from the other:
- It separately models the "bump size" and the "readout resolution"
  rather than using the 1/cellsPerModuleAxis for both.
- It models how a union of bumps is harder to readout than a single
  bump. (In this approach, more cells become "active" than just the
  union of the active cells)
- Hexagonal grids. The firing fields of a cell are arranged in a
  hexagon rather than in a square.
- Each firing field is round instead of square.

Other changes:

- Fix a bug: PIUNExperiment.representationSet wasn't being properly
  filled, so we weren't correctly detecting temporary convergences on
  a wrong object.
- Stop seeding the object generator with the number of objects. It
  should generate different objects on different runs.
- Add couple other capabilities to convergence simulation.
- Tweak the Superficial2DLocationModule so that it shares a more
  common interface with the new ThresholdedGaussian2DLocationModule.