Method for selecting task network, system and method for determining actions based on sensing data

A method for determining actions based on sensing data, comprising:

• obtaining a plurality of first individuals that belong to a first population and a plurality of second individuals that belong to a second population, wherein the plurality of first individuals and the plurality of second individuals are evolved via a coevolution process, wherein the coevolution process comprises one of the following:
• a cooperative coevolution and a competitive coevolution;
• obtaining a plurality of task networks generated by a plurality of pattern-producing networks, wherein each task network of the plurality of task networks includes a first network layer and a second network layer, the first network layer and the second network layer include a plurality of nodes, and the plurality of nodes of the first network layer is connected to the plurality of nodes of the second layer based on a weight value initialized by a corresponding pattern-producing network;
• obtaining sensing data and mapping the sensing data into an input feature vector, wherein the sensing data comprise a plurality of sensing attributes that belong to a first specific individual, and the first specific individual belongs to the first population;
• feeding the input feature vector into a specific task network among the plurality of task networks, so that the specific task network generates an output feature vector, wherein the specific task network is trained based on the first population and the second population, wherein the weight value of the specific task network is adjusted during training based on the plurality of first individuals and the plurality of second individuals that are coevolved, wherein the specific task network corresponds to a specific pattern-producing network with highest fitness score among the plurality of pattern-producing networks, wherein the fitness score of a pattern-producing network of the plurality of pattern-producing networks is determined based on an accuracy of the each task network that corresponds to the pattern-producing network, wherein the accuracy is evaluated after training the each task network;
• and
• determining an output action according to the output feature vector, and setting a second specific individual based on the output action, wherein the second specific individual belongs to the second population.

A system that determines actions based on sensing data, comprising:

• a sensing circuit;
• a communication circuit;
• a storage circuit, recording a plurality of modules;
• and
• a processor, coupled to the storage circuit, and accessing the plurality of modules to execute the following:
• obtaining a plurality of first individuals that belong to a first population and a plurality of second individuals that belong to a second population, wherein the plurality of first individuals and the plurality of second individuals are evolved via a coevolution process, wherein the coevolution process comprises one of the following:
• a cooperative coevolution and a competitive coevolution;
• obtaining a plurality of task networks generated by a plurality of pattern-producing networks, wherein each task network of the plurality of task networks includes a first network layer and a second network layer, the first network layer and the second network layer include a plurality of nodes, and the plurality of nodes of the first network layer is connected to the plurality of nodes of the second layer based on a weight value initialized by a corresponding pattern-producing network;
• obtaining sensing data from the sensing circuit, and mapping the sensing data into an input feature vector, wherein the sensing data comprise a plurality of sensing attributes that belong to a first specific individual, and the first specific individual belongs to the first population;
• feeding the input feature vector into a specific task network among the plurality of task networks so that the specific task network generates an output feature vector, wherein the specific task network is trained based on the first population and the second population, and the weight value of the specific task network is adjusted during training based on the plurality of first individuals and the plurality of second individuals that are coevolved, wherein the specific task network corresponds to a specific pattern-producing network with highest fitness score among the plurality of pattern-producing networks, wherein the fitness score of a pattern-producing network of the plurality of pattern-producing networks is determined based on an accuracy of the each task network that corresponds to the pattern-producing network, wherein the accuracy is evaluated after training the each task network;
• and
• determining an output action according to the output feature vector, and controlling the communication circuit to send the output action to a second specific individual that belongs to the second population, wherein the output action is configured to set the second specific individual.