17686147. COMPUTE-IN-MEMORY SYSTEMS AND METHODS WITH CONFIGURABLE INPUT AND SUMMING UNITS simplified abstract (TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, LTD.)
COMPUTE-IN-MEMORY SYSTEMS AND METHODS WITH CONFIGURABLE INPUT AND SUMMING UNITS
Organization Name
TAIWAN SEMICONDUCTOR MANUFACTURING COMPANY, LTD.
Inventor(s)
Chia-En Huang of Hsinchu County (TW)
Yi-Ching Liu of Hsinchu City (TW)
Wen-Chang Cheng of Richmond TX (US)
COMPUTE-IN-MEMORY SYSTEMS AND METHODS WITH CONFIGURABLE INPUT AND SUMMING UNITS - A simplified explanation of the abstract
This abstract first appeared for US patent application 17686147 titled 'COMPUTE-IN-MEMORY SYSTEMS AND METHODS WITH CONFIGURABLE INPUT AND SUMMING UNITS
Simplified Explanation
The abstract of the patent application describes a device that consists of a multiplication unit and a configurable summing unit. The multiplication unit receives data and weights for a specific layer and performs multiplication operations to generate multiplication results. The configurable summing unit, on the other hand, is configured by layer values and performs addition operations on a specific number of inputs. It then sums the multiplication results and provides an output.
- The device includes a multiplication unit and a configurable summing unit.
- The multiplication unit multiplies data by weights to produce multiplication results.
- The configurable summing unit receives layer values and performs addition operations on a specific number of inputs.
- The summing unit then sums the multiplication results and provides an output.
Potential Applications:
- Artificial neural networks
- Deep learning algorithms
- Image and speech recognition systems
- Natural language processing
Problems Solved:
- Efficient processing of large amounts of data in neural networks
- Configurable and adaptable architecture for different layers
- Simplified implementation of complex mathematical operations
Benefits:
- Improved performance and accuracy in machine learning tasks
- Flexibility to adjust the configuration for different layers
- Reduced computational complexity and power consumption
Original Abstract Submitted
A device includes a multiplication unit and a configurable summing unit. The multiplication unit is configured to receive data and weights for an Nth layer, where N is a positive integer. The multiplication unit is configured to multiply the data by the weights to provide multiplication results. The configurable summing unit is configured by Nth layer values to receive an Nth layer number of inputs and perform an Nth layer number of additions, and to sum the multiplication results and provide a configurable summing unit output.