You are here

April release

April's release is here. I've been doing a great deal of work on making it easier to swap out associative learning systems, so most of the fixes and features are related to that.


  1. IAssociativeLinkageSystem now provides the parameter handler so that links can be (de)serialized correctly, regardless of new link parameters
  2. IAssociativeLink is now fully IParameterized (making it easier to probe)
  3. ISubsymbolicChunk.setSourceActivation() has been changed so that it tracks each source of activation (i.e., the buffer) instead of relying upon the ISourceActivationSpreader to handle activation from multiple buffers
  4. Fixed some bugs associated with the merging of associative links
  5. Fixed a bug in the AbstractActivationBuffer that would prevent it from adding a chunk if an identical one was already in the buffer


  1. Added support for probing the parameters of associative links


  1. The associative link viewer now uses curved connections so that bidirectional links don't clobber each other (see below)
  2. The general probe viewer can now plot out associative link parameters (see below)

As always, if you see some behavior that you think isn't right, let me know.


Better associative link viewing

Graphing associative parameters (yes, I'm working on associative learning)

This is an example probe configuration file for the associative links. It probes on the model named "participant", and only those j chunks that are perceptual-symbols named either "tunnel" or "cowboy". Furthermore, it only displays the strengths of those links that are to i chunks that are of the chunk-type concept-pair.

<group id="strength">
  <model pattern="participant">
   <chunk-type pattern="perceptual-symbol"> 
    <chunk pattern="(tunnel)|(cowboy)">
     <link chunk=".*" chunk-type="concept-pair">
       <probe pattern="Strength" poll="true" />