Sketch of conceptual system before diving deeper into LIDA.

Posted: November 1, 2011 at 4:39 pm

In this post I will describe the conceptual subsystem, as I have imagined it up to this point, before looking more closely into the LIDA architecture. The central feature of this conception is that each high level feature, for example location in space, or colour (or redness), is considered independent. This contrasts with many systems that concatenate all these features in order to create a cluster representation that encapsulates overall similarity. This is conceptualized as a highly shallow system where features are meant to be high level. In fact no clustering is needed at all, as only a distance function can be used to determine the distance between precepts in terms of each individual feature.

As each of these features is independent, the neuron like units (NLUs) that represent percepts, have links corresponding to each feature. Each link would then have a different weight, correlating to the similarity of those precepts. Activation would pass through all links simultaneously, and neural inspired mechanisms such as inhibition could constrain activation. It is expected that the continued activation of two precepts will increase the connection strength between them. Below is a sample activation: Percept A can activate Percept B in terms of spatial location, which can activate Percept C in terms of colour similarity. Of course, as activation is passed through all links a clean association like this is unlikely. Perhaps the strongest link between two precepts is strengthened, while the others are weakened. Over time this could limit the flexibility of associations, or could lead to stable patterns of activation that could be something like high level concepts. A central issue is how to limit complexity of the network. One idea I posted here was that SWS sleep could prune weak links, and REM sleep could strengthen remaining links.