Abstract
From the background information supplied by the inventors, news correspondents o btained the followingquote: “In order to perform machine learning in hardware t he desired output is to be achieved from aparticular set of input data. For exa mple, input data (e.g. an input vector) is provided to a first layer.The input data is multiplied by a matrix of values, or weights, for the layer. The output signals (or outputvector) for the layer are the result of the matrix multiplica tion in the layer. The output signals are providedas the input signals to the n ext layer of matrix multiplications. This process may be repeated for a largenu mber of layers, each of which may include a number of neurons. The final output signals of the lastlayer are desired to match a particular set of target values . To perform machine learning, the weights (e.g.resistances) for one or more of the layers are adjusted in order to bring the final output signals closer toth e target values. Although this process can theoretically alter the weights of th e layers to provide thetarget output, in practice, ascertaining the appropriate set of weights is challenging. Various techniquesexist in order to aid in dete rmining the weights. However, each of these techniques also face significantiss ues.”
NETL
NSTL