18518151. SRAM ARCHITECTURE FOR CONVOLUTIONAL NEURAL NETWORK APPLICATION simplified abstract (Taiwan Semiconductor Manufacturing Company, Ltd.)
SRAM ARCHITECTURE FOR CONVOLUTIONAL NEURAL NETWORK APPLICATION
Organization Name
Taiwan Semiconductor Manufacturing Company, Ltd.
Inventor(s)
Jaw-Juinn Horng of Hsinchu (TW)
SRAM ARCHITECTURE FOR CONVOLUTIONAL NEURAL NETWORK APPLICATION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18518151 titled 'SRAM ARCHITECTURE FOR CONVOLUTIONAL NEURAL NETWORK APPLICATION
Simplified Explanation
The abstract describes a convolutional neural network (CNN) that includes memory cells with capacitive elements to generate products based on weight and input bits, enabling the elements when a threshold is met. The network also includes a reference cell array and a memory controller to determine element activation based on signal comparisons.
- Memory cells in the CNN have capacitive elements that multiply weight and input bits to produce a product.
- The capacitive elements are enabled when the product meets a predetermined threshold.
- A reference cell array with capacitive elements and a memory controller are used to compare signals and determine element activation.
Potential Applications
- Image recognition
- Speech recognition
- Autonomous driving systems
Problems Solved
- Efficient processing of large amounts of data
- Improved accuracy in pattern recognition tasks
Benefits
- Reduced power consumption
- Faster processing speeds
- Enhanced performance in complex tasks
Original Abstract Submitted
One aspect of this description relates to a convolutional neural network (CNN). The CNN includes a memory cell array including a plurality of memory cells. Each memory cell includes at least one first capacitive element of a plurality of first capacitive elements. Each memory cell is configured to multiply a weight bit and an input bit to generate a product. The at least one first capacitive element is enabled when the product satisfies a predetermined threshold. The CNN includes a reference cell array including a plurality of second capacitive elements. The CNN includes a memory controller configured to compare a first signal associated with the plurality of first capacitive elements with a second signal associated with at least one second capacitive element of the plurality of second capacitive elements, and, based on the comparison, determine whether the at least one first capacitive element is enabled.
