Aggregate adjustments in a cross bar neural network
AssigneeInternational Business Machines Corporation
InventorsEffendi Leobandung, Zhibin Ren, SEYOUNG KIM, Paul Michael Solomon
Life of this patent
Estimating the term…
Abstract
Method, systems, crosspoint arrays, and systems for tuning a neural network. A crosspoint array includes: a set of conductive rows, a set of conductive columns intersecting the set of conductive rows to form a plurality of crosspoints, a circuit element coupled to each of the plurality of crosspoints configured to store a weight of the neural network, a voltage source associated with each conductive row, a first integrator attached at the end of at least one of the conductive column, and a first variable resistor attached to the integrator and the end of the at least one conductive column.
Claims20
A method comprising:
- applying a voltage to each one of a plurality of conductive rows of a neural network, wherein the plurality of conductive rows intersect a plurality of conductive columns to form a plurality of crosspoints, and wherein a circuit element at each of the plurality of crosspoints is configured to store a weight of the neural network;
- and
- determining at least one of a reference voltage and a reference current for each of a plurality of variable resistors, wherein each of the plurality of variable resistors and each of a plurality of integrators are connected to a respective end of one of the plurality of conductive columns.
A computer program product for determining a parameter of a neural network circuit, the computer program product comprising:
- a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to:
- apply a voltage to each one of a plurality of conductive rows of a neural network, wherein the plurality of conductive rows intersect a plurality of conductive columns to form a plurality of crosspoints, and wherein a circuit element at each of the plurality of crosspoints is configured to store a weight of the neural network;
- and
- determine at least one of a reference voltage and a reference current for each of a plurality of variable resistors, wherein each of the plurality of variable resistors and each of a plurality of integrators are connected to a respective end of one of the plurality of conductive columns.
A crosspoint array for implementing a neural network, the crosspoint array comprising:
- a set of conductive rows;
- a set of conductive columns intersecting the set of conductive rows to form a plurality of crosspoints;
- a circuit element coupled to each of the plurality of crosspoints configured to store a weight of the neural network;
- a voltage source associated with each conductive row;
- a first integrator attached at the end of at least one of the conductive column;
- and
- a first variable resistor attached to the integrator and the end of the at least one conductive column.
Dependencies are inferred from the claim text.