View source for 17847955. CROSSBAR ARRAYS IMPLEMENTING TRUTH TABLES simplified abstract (INTERNATIONAL BUSINESS MACHINES CORPORATION)

=CROSSBAR ARRAYS IMPLEMENTING TRUTH TABLES=


==Organization Name==

[[:Category:INTERNATIONAL BUSINESS MACHINES CORPORATION|INTERNATIONAL BUSINESS MACHINES CORPORATION]]


[[Category:INTERNATIONAL BUSINESS MACHINES CORPORATION]]

==Inventor(s)==

[[:Category:Charles Mackin of San Jose CA (US)|Charles Mackin of San Jose CA (US)]][[Category:Charles Mackin of San Jose CA (US)]]

==CROSSBAR ARRAYS IMPLEMENTING TRUTH TABLES - A simplified explanation of the abstract==


This abstract first appeared for US patent application 17847955 titled 'CROSSBAR ARRAYS IMPLEMENTING TRUTH TABLES

==Simplified Explanation==

The abstract describes a method for preparing a trained crossbar array of a neural network. Here is a simplified explanation of the abstract:

* The method involves using a computer simulation of a crossbar array to train the neural network.
* A predetermined truth table is used as input, and analog output values are generated based on simulated weights.
* Loss values are calculated by comparing the analog output values with expected values for the output portion of the truth table.
* The simulated weights are adjusted based on the calculated loss values.
* The input portion of the truth table is repeatedly fed into the simulation and the output values are recalculated using the adjusted weights.
* This process continues until the analog output values match the expected values within a predefined margin of error.

Potential applications of this technology:

* Artificial intelligence and machine learning systems
* Pattern recognition and image processing
* Natural language processing and speech recognition
* Autonomous vehicles and robotics
* Financial analysis and prediction

Problems solved by this technology:

* Efficient training of neural networks using a computer simulation
* Optimization of weights in a crossbar array for accurate output prediction
* Reducing the margin of error in neural network predictions

Benefits of this technology:

* Faster and more efficient training of neural networks
* Improved accuracy and reliability of neural network predictions
* Cost-effective implementation of neural network systems
* Scalability for handling large datasets and complex problems


==Original Abstract Submitted==

A method for preparing a trained crossbar array of a neural network is provided. The method includes feeding an input portion of a predetermined truth table into a computer simulation of a crossbar array, and generating analog output values for the input portion of the truth table based on simulated weights. The method further includes calculating a loss value from each of the analog output values and expected values for an output portion of the truth table, and adjusting the simulated weights based on the calculated loss values. The method further includes refeeding the input portion of the predetermined truth table into the computer simulation and recalculating the output values using the adjusted simulated weights until the analog output values produce the expected values for the output portion of the truth table within a predefined margin of error.



[[Category:G06N3/063]]
[[Category:G06N3/08]]