This week on Journal Club session Aamir Khan will talk about a paper "Study of weak / strong multiplication in gain modulation".
This is an extension of my previous work on the multiplicative response of spiking neural networks. In our artificial life platform I evolve a gain modulated network where there are two Inputs namely d (driving input) and m ( the modulatory input). The network aims for a multiplicative Sigmoidal/gaussian response modulated by the modulatory input m. The response of the network exhibits three properties such as 1. Saturation , 2. Non linearity and 3.Multiplicity. The multiplicative capacity of the network varies from strong to weak in a manner that for some modulatory input there is strong multiplication and for some there is a weak multiplication, I have differentiated the response of the network in weak/ strong response. Along with the classification some further constraints are introduced in the system like The size of the network tends to be larger i.e 5 nodes, in order to reduce the size I reward the response of the one interneuron and treat it as the output node.
- Steuber, Volker and Davey, Neil and Wrobel, Borys. (2018). "Spiking Neural Networks Evolved to Perform Multiplicative Operations" , 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018, Proceedings, Part I. 10.1007/978-3-030-01418-6_31.