| 主权项 |
1. A computer-based multi-layer artificial network named continuous-weight neural network (CWNN) configured to receive an input feature set wherein the input feature set comprises a variable number of features wherein each feature of the variable number of features is represented by an input pair comprising an input value and an input point provided to the CWNN, the input value being a real number input value and the input point being a real vector input point, the CWNN further comprising:
a main network and a control network,
wherein the control network is a multilayer perceptron which comprises: i) an input layer comprising CI nodes, and wherein CI equals to a dimension of the real vector input point and ii) an output layer comprising CO nodes, andwherein the main network comprises: i) a hidden layer comprising CO nodes and ii) a plurality of input nodes equal to the variable number of features, wherein the input value of the variable number of features is fed to the input nodes; CO connections between the CO nodes of the hidden layer and an input node of the main network; and CO weights associated to each of the CO connections, wherein the CO weights are in correspondence of an output of the control network; wherein the output of the control network is in correspondence of an input point vector associated to an input value fed to the input node of the main network, and a set comprising the CO weights is equal to the output of the control network, the output of the control network being based on the input point fed to an input layer of the control network. |
