Analysis of Adaptivity and Plasticity in the Network of Homeostatic Neurons
Nikitin O.Yu., Lukyanova O.A., Kunin A.S.
Computing Center FEB RAS, Khabarovsk, Russia
Khabarovsk State University of Economics and Law, Khabarovsk, Russia
Abstract. Here we present the model of bioinspired neuron, incorporating cellular homeostasis. We propose the modified STDP plasticity rule, including influence of the input signal on the neuronal cellular homeostatic state. Properties of the network of proposed neurons have been studied. The network was used to control adaptive behavior of an agent in the multi-task environment. Statistical study was performed to find out the adaptive properties of the network with different levels of homeostatical plasticity. Model of the network, incorporating full homeostatically dependent weight correction was able to control an agent in the environment without direct evaluation function. We conclude that important to pay attention to interaction of the input signal and cellular homeostasis in order to reproduce biologically plausible and adaptive behavior of an agent.
Key words: cellular homeostasis, neuronal plasticity, adaptive behavior.