Analog neurons and neurosynaptic networks

A neural network comprising:

• one or more weighting units, each of said weighting units having a weighting factor, said weighting factor being substantially determined by the resistance of said weighting unit;
• and
• one or more accumulation units, each of said accumulation units comprising a phase change material, said phase change material having a plurality of states, said states including:
• a reset state;
• a set state obtainable from said reset state by applying energy in an amount corresponding to the set energy of said reset state, said set energy corresponding to the threshold of said accumulation unit, said set state having lower resistance than said reset state, said set state being transformable to said reset state by applying energy in an amount corresponding to the reset energy of said set state;
• and one or more accumulation states having substantially the same resistance as said reset state, said accumulation states being obtainable from said reset state by applying energy in an amount less than said threshold;
• wherein each of said weighting units is connected in series with at least one of said accumulation units, each of said accumulation units is connected in series with at least one of said weighting units, said phase change material of each of said accumulation units is in said reset state or one of said accumulation states, and said weighting factors determine the extent of electrical communication between the interconnected of said weighting units and said accumulation units.

A neural network comprising:

• data input means, said data input means producing two or more signals;
• and
• an analog neuron, said analog neuron comprising a phase change material, said phase change material having a plurality of states, said states including:
• a reset state;
• a set state obtainable from said reset state by applying energy in an amount corresponding to the set energy of said reset state, said set energy corresponding to the threshold of said analog neuron, said set state having lower resistance than said reset state, said set state being transformable to said reset state by applying energy in an amount corresponding to the reset energy of said set state;
• and one or more accumulation states having substantially the same resistance as said reset state, said accumulation states being obtainable from said reset state by applying energy in an amount less than said threshold;
• wherein at least two of said two or more signals are provided as input signals to said analog neuron, said analog neuron cumulatively responding to said input signals by transforming among said accumulation states, said analog neuron firing when said input signals provide energy in an amount sufficient to set said phase change material.