Experimenting with human-like artificial Intelligence

   Dr. M.Wagaarachchi

Humans, with our sophisticated Neocortex, are known to have a six-layer neural network to perform complex brain functions. Artificial Neural Networks (ANNs) are devised to closely replicate the problem-solving abilities of the human brain.

           Prof. Asoka

Hasitha set out to understand more about hidden layers, perhaps at the cost of performance, because of the associated complexity in training and usage. But what is the optimum number of hidden layers needed to solve a particular problem? That was Mihirini’s research question.

Mimicking neuroplasticity of the human brain - which enables it to reorganize its neural connections - Mihirini designed and developed an algorithm for her investigations. She applied this to ANNs with upto twelve hidden layers, and observed that a simpler architecture corresponding to a given complex one can be generated through her approach.


An Initiative by MoraMinds                                                                                           1 this post on 1