An improved continuous Hopfield neural network: computing LS parameters and lost function as well as multi-order identification

We focus on an improved structure for the continuous Hopfield neural network (CHNN) in order to give an output of least square (LS) parameters and its corresponding lost function simultaneously. Based on the structure and by controlling the connection status of the corresponding feedback loop of the network on and off, we can compute the LS parameters and lost functions for each LS model order which is lower than the input matrix. Also the improved structure can make the CHNN´s fixed structure adapt to any input matrix which may have dimensions lower than or equal to the network´s dimensions. Simulation results show that the improved structure can support the functions we have given