Method of neural network construction for the simulation of physical systems

A method of data compression, comprising:

• constructing a neural network comprising a compression block receiving a datum X as input and a decompression block generating the datum X as output;
• the neural network comprising processing nodes and connections between the processing nodes forming a topology organized in layers, such that each layer is defined by a set of simultaneously calculable processing nodes, and an input of a processing node of a layer is connectable to an output of a processing node of any previously calculated layer;
• wherein constructing the neural network includes initializing the neural network according to an initial topology comprising an input layer, at least one hidden layer comprising at least one node, and a set of output nodes, and implementing at least one topological optimization phase on the compression block and the decompression block, each topological optimization phase comprising:
• at least one additive phase comprising a modification of the network topology by adding at least one processing node and/or a connection link between the input of a processing node of a layer and the output of a processing node of any layer;
• and/or
• at least one subtractive phase comprising a modification of the network topology by removing at least one processing node and/or removing a connection link between processing nodes of two layers,
• wherein each topology modification comprises a selection of a topology modification among several candidate modifications, based on an estimation of the variation in the network error, calculated on training data, between a previous topology and each topology modified according to a candidate modification and
• applying the compression block of the constructed network to at least one datum representative of the state of a physical system.