Reconfigurable computing architecture for implementing artificial neural networks

A computer (CALC) for computing a layer (Ck, Ck+1) of an artificial neural network, the neural network being formed of a sequence of layers (Ck, Ck+1) each consisting of a set of neurons, each layer being associated with a set of synaptic coefficients (wi,jk+1) forming at least one weight matrix ([MP]k+1, WP,Q), the computer (CALC) being able to be configured in accordance with two separate configurations (CONF1, CONF2) and comprising:

• a transmission line (L_data) for distributing input data (Xjk, δik+1, xi,j),
• a set of computing units (PE0, PE1, PE2, PE3) of ranks n=0 to N, where N is an integer greater than or equal to 1, for computing an input data sum weighted by synaptic coefficients,
• a set of weight memories (MEM_POIDS0, MEM_POIDS1, MEM_POIDS2, MEM_POIDS3) each associated with a computing unit (PE0, PE1, PE2, PE3), each weight memory containing a subset of synaptic coefficients required and sufficient for the associated computing unit (PE0, PE1, PE2, PE3) to carry out the computations necessary for either one of the two configurations (CONF1, CONF2),
• control means for configuring the computing units (PE0, PE1, PE2, PE3) of the computer (CALC) in accordance with either one of the two configurations (CONF1, CONF2),
• in the first configuration (CONF1), the computing units being configured such that a weighted sum is computed in full by one and the same computing unit,
• in the second configuration (CONF2), the computing units being configured such that a weighted sum is computed by a chain of multiple computing units arranged in series.