18489210. DEVICE AND METHOD FOR PROCESSING A CONVOLUTIONAL NEURAL NETWORK WITH BINARY WEIGHTS simplified abstract (Huawei Technologies Co., Ltd.)

From WikiPatents
Jump to navigation Jump to search

DEVICE AND METHOD FOR PROCESSING A CONVOLUTIONAL NEURAL NETWORK WITH BINARY WEIGHTS

Organization Name

Huawei Technologies Co., Ltd.

Inventor(s)

Van Minh Nguyen of Boulogne Billancourt (FR)

DEVICE AND METHOD FOR PROCESSING A CONVOLUTIONAL NEURAL NETWORK WITH BINARY WEIGHTS - A simplified explanation of the abstract

This abstract first appeared for US patent application 18489210 titled 'DEVICE AND METHOD FOR PROCESSING A CONVOLUTIONAL NEURAL NETWORK WITH BINARY WEIGHTS

Simplified Explanation

The abstract of this patent application describes various embodiments related to convolutional neural networks (CNNs) that use a convolution kernel with binary weights. The CNN is trained with this convolution kernel to determine a set of binary weights, which are then used for inference.

  • CNNs with convolution kernels configured with binary weights are disclosed.
  • The CNN is trained using the convolution kernel to determine a set of binary weights.
  • The set of binary weights is used for inference of the CNN.
  • The patent application covers devices, methods, and computer programs related to this technology.

Potential Applications:

  • Image recognition and classification: CNNs with binary weights can be used for tasks such as object recognition, facial recognition, and image classification.
  • Natural language processing: CNNs can be applied to tasks like sentiment analysis, text classification, and language translation.
  • Autonomous vehicles: CNNs with binary weights can be used for object detection and recognition in self-driving cars.

Problems Solved:

  • Improved efficiency: Using binary weights in CNNs can reduce memory requirements and computational complexity, making them more efficient for deployment on resource-constrained devices.
  • Faster inference: Binary weights can enable faster inference times, allowing real-time processing of data in applications like autonomous vehicles or real-time video analysis.

Benefits:

  • Reduced memory footprint: Binary weights require less memory compared to traditional floating-point weights, making CNNs more suitable for deployment on devices with limited memory.
  • Lower power consumption: The reduced computational complexity of binary weights can lead to lower power consumption, making CNNs more energy-efficient.
  • Faster inference: Binary weights enable faster inference times, improving the responsiveness of applications that rely on real-time processing.

Overall, this patent application describes a method to train CNNs using convolution kernels with binary weights, offering potential applications in image recognition, natural language processing, and autonomous vehicles. The technology solves problems related to efficiency and inference speed, while providing benefits such as reduced memory footprint, lower power consumption, and faster inference times.


Original Abstract Submitted

Various embodiments relate to convolutional neural networks (CNN). CNN may be provided with a convolution kernel configured with binary weights. The CNN may be trained with the convolution kernel to determine a set of binary weights for the convolution kernel. The set of binary weights may be used for inference of the CNN. Devices, methods, and computer programs are disclosed.

Retrieved from "http://wikipatents.org/index.php?title=18489210._DEVICE_AND_METHOD_FOR_PROCESSING_A_CONVOLUTIONAL_NEURAL_NETWORK_WITH_BINARY_WEIGHTS_simplified_abstract_(Huawei_Technologies_Co.,_Ltd.)&oldid=33539"