Electrical circuit design

A machine learning system for designing an electrical circuit, the machine learning system comprising:

• a neural network having a plurality of network parameter values, the neural network being adapted to receive a state signal representing a state of an environment and to process the state signal according to the plurality of network parameter values to generate action selection data representing an action selection policy for selecting an action to be performed by an agent in response to the state signal;
• a buffer;
• an interface for receiving input data representing three or more circuit elements in the electrical circuit, the three or more circuit elements including a first circuit element and a second circuit element;
• a processing engine adapted to:
• process the input data to generate a representation of the electrical circuit, the representation being adapted to generate a state signal representing a configuration of the electrical circuit in response to receiving an action selection signal representing a decision with respect to a configuration of the electrical circuit;
• identifying a plurality of candidate routes for connecting the first circuit element to the second circuit element based on the representation, each of the plurality of candidate routes comprising a respective set of candidate sub-routes;
• selecting a candidate route from the plurality of candidate routes by iteratively performing the steps of:
• using a Sequential Monte Carlo process to perform a look ahead search of a subset of the plurality of candidate routes by generating simulations in respect of candidate sub-routes of the subset of the plurality of candidate routes and according to the configuration, wherein the Sequential Monte Carlo process is guided using a neural network;
• determining target action selection data associated with the configuration by evaluating the simulations;
• updating the representation of the electrical circuit by providing an action selection signal representing a selection of a candidate sub-route based on the determined target action selection data;
• and
• receiving a state signal representing a configuration of the representation of the electrical circuit.