Recent Posts

• December 22, 2019

  Similarity-driven Brain

  Human brain has so various functions that we couldn’t solve the whole mystery yet. In order to understand the function of brain or its nature per se, people have suggested and developed many computational models and explained our brain. This is based on the basic assumption that human brain internally computes or handles symbols or representations of external worlds. The Artificial Neural Network (ANN) is a way of doing that and it has many industrial applications in wide ranges, as well as in brain research area. The ANN, however, is successful only in small problem-specific areas such as pattern recognition, classification, associated memory, etc. In this work, I explored the nature of brain more deeply and made a general assumption that human brain always seeks similarity among external stimulus. On this assumption, I suggest a general neural network model, which is more similar to real neurons in brain than other ANNs, and is suitable for large networks due to its scale-free feature. I also suggest that basic functions, such as logical operation and classification, can also be implemented by seeking similarity among stimulus. The suggested model is simulated in Matlab® environment and shows expected behaviors, though there still remain some issues to generalize it.

• May 07, 2019

  Topological exploration on spiking patterns of hierarchically organized Izhikevich’s neuron model

  In a complex system like brain, it is hard to determine the factor to influence the global spiking patterns. One of such factors, in this work, we explored spiking patterns of hierarchically organized neurons. For simulations, we introduced Izhikevich’s neuron model, which enabled us to simulate large-scale networks in a relevant time. To modulate hierarchical organizations, we controlled the hierarchical levels and the number of sub-modules in a module. In addition, we explored the effects of edge density and node degree in a given network. The result showed that neural activities are dependent on the hierarchical organizations and it evokes different spiking patterns for small and large networks. And, it is also important to control the number of synaptic connections in a node so that limited sustained activity may be maintained. Therefore, we suggest that biologically plausible model such as Izhikevich’s model is helpful to study hierarchical organizations optimized for large-scale neural networks.

• April 18, 2019

  Rethinking phonological loop in working memory model: phonological divergence of the direct pathway

  Introduction. Most cognitively demanding tasks require temporal storages of transient information, i.e. short-term store (STS). It is located between sensory stores and long-term stores, and used to hold and manipulate information for cognitive tasks as a working memory1. Though the dual-task study2 showed that STS was really acting as a working memory, the processing time and accuracy weren’t enough to account for working memory. As a result, it was assumed that the working memory is not simply a singular storage unit like STS, but made up of a few components.