18449200. INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND NON-TRANSITORY COMPUTER-READABLE STORAGE MEDIUM simplified abstract (CANON KABUSHIKI KAISHA)
INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND NON-TRANSITORY COMPUTER-READABLE STORAGE MEDIUM
Organization Name
Inventor(s)
INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND NON-TRANSITORY COMPUTER-READABLE STORAGE MEDIUM - A simplified explanation of the abstract
This abstract first appeared for US patent application 18449200 titled 'INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, AND NON-TRANSITORY COMPUTER-READABLE STORAGE MEDIUM
Simplified Explanation
The patent application describes an information processing apparatus that includes a conversion unit and an operation unit for processing weight coefficients in neural networks.
- Conversion unit outputs weight coefficients as is if they are N bits wide, otherwise converts them to N bits.
- Operation unit performs multiply-accumulate operations between weight coefficients and feature images to generate output images.
- Potential Applications:**
- Image recognition systems
- Speech recognition systems
- Natural language processing systems
- Problems Solved:**
- Efficient processing of weight coefficients in neural networks
- Improved accuracy in feature image generation
- Benefits:**
- Faster processing speeds
- Reduced computational resources required
- Enhanced performance in neural network applications
Original Abstract Submitted
An information processing apparatus comprises a conversion unit configured to, if a bit width of a weight coefficient in a first layer in weight coefficients in a plurality of layers is N bits, wherein N is a predetermined integer not less than 2, output the weight coefficient, and if the bit width of the weight coefficient is not N bits, output a weight coefficient of N bits obtained by converting the weight coefficient, and an operation unit configured to generate a feature image in a second layer based on a multiply-accumulate operation between the weight coefficient output by the conversion unit and a feature image in the first layer.
