18469272. GENERAL PADDING SUPPORT FOR CONVOLUTION ON SYSTOLIC ARRAYS simplified abstract (GOOGLE LLC)
Contents
GENERAL PADDING SUPPORT FOR CONVOLUTION ON SYSTOLIC ARRAYS
Organization Name
Inventor(s)
David Alexander Majnemer of Mountain View CA (US)
Blake Alan Hechtman of Mountain View CA (US)
Bjarke Hammersholt Roune of Mountain View CA (US)
GENERAL PADDING SUPPORT FOR CONVOLUTION ON SYSTOLIC ARRAYS - A simplified explanation of the abstract
This abstract first appeared for US patent application 18469272 titled 'GENERAL PADDING SUPPORT FOR CONVOLUTION ON SYSTOLIC ARRAYS
The abstract of this patent application describes methods and systems for performing convolutional computations for a neural network on a hardware circuit with a matrix computation unit.
- Receiving a request to perform convolutional computations on a feature tensor and a filter with padding applied.
- Generating instructions for the hardware circuit to transfer data from main memory to scratchpad memory.
- Repeatedly identifying subsets of the feature tensor and checking consistency between scratchpad memory and main memory views.
Potential Applications: - This technology can be applied in various fields such as image recognition, natural language processing, and autonomous vehicles.
Problems Solved: - Efficiently performing convolutional computations on neural networks. - Optimizing memory usage for large-scale computations.
Benefits: - Faster processing of neural networks. - Reduced memory overhead. - Improved performance in complex tasks.
Commercial Applications: - This technology can be utilized in industries such as healthcare for medical image analysis, in finance for fraud detection, and in retail for customer behavior analysis.
Questions about Convolutional Computations: 1. How does this technology improve the efficiency of convolutional computations?
- This technology optimizes memory usage and speeds up the processing of neural networks, leading to improved performance in various applications.
2. What are the potential commercial uses of this innovation?
- The commercial applications of this technology span across industries such as healthcare, finance, and retail, where it can enhance tasks like medical image analysis, fraud detection, and customer behavior analysis.
Original Abstract Submitted
Methods and systems, including computer programs encoded on a computer storage medium. In one aspect, a method includes the actions of receiving a request to perform convolutional computations for a neural network on a hardware circuit having a matrix computation unit, the request specifying the convolutional computation to be performed on a feature tensor and a filter and padding applied to the feature tensor prior to performing the convolutional computation; and generating instructions that when executed by the hardware circuit cause the hardware circuit to perform operations comprising: transferring feature tensor data from a main memory of the hardware circuit to a scratchpad memory of the hardware circuit; and repeatedly performing the following operations: identifying a current subset of the feature tensor; and determining whether a memory view into the scratchpad memory for the current subset is consistent with a memory view of the current subset in the main memory.
