Google llc (20240346298). SYSTEM AND METHOD FOR PARALLELIZING CONVOLUTIONAL NEURAL NETWORKS simplified abstract

SYSTEM AND METHOD FOR PARALLELIZING CONVOLUTIONAL NEURAL NETWORKS

Organization Name

google llc

Inventor(s)

Alexander Krizhevsky of Mountain View CA (US)

Ilya Sutskever of San Francisco CA (US)

Geoffrey E. Hinton of Toronto (CA)

SYSTEM AND METHOD FOR PARALLELIZING CONVOLUTIONAL NEURAL NETWORKS - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240346298 titled 'SYSTEM AND METHOD FOR PARALLELIZING CONVOLUTIONAL NEURAL NETWORKS

The abstract describes a patent application for a parallel convolutional neural network implemented by multiple CNNs on separate processing nodes, with interconnected layers for forward activation propagation.

• Each processing node houses a convolutional neural network (CNN) with multiple layers.
• Some layers within each CNN are interconnected across processing nodes to feed activations forward.
• Other layers within each CNN are not interconnected across nodes.
• This design allows for parallel processing and efficient utilization of resources.
• The innovation aims to improve the performance and scalability of convolutional neural networks.

Potential Applications:

• Image recognition and classification tasks.
• Natural language processing applications.
• Autonomous driving systems.
• Medical image analysis.
• Video processing and analysis.

Problems Solved:

• Enhancing the speed and efficiency of CNNs.
• Facilitating parallel processing for large-scale data sets.
• Improving the accuracy of deep learning models.
• Addressing resource constraints in neural network training.
• Enabling real-time decision-making in various applications.

Benefits:

• Faster processing of complex data.
• Enhanced accuracy in pattern recognition.
• Scalability for handling large volumes of data.
• Improved performance in deep learning tasks.
• Real-time decision-making capabilities.

Commercial Applications:

• Cloud computing services.
• Edge computing devices.
• Robotics and automation systems.
• Surveillance and security systems.
• Healthcare diagnostics and monitoring.

Questions about Parallel Convolutional Neural Networks: 1. How does the parallel architecture of CNNs improve processing speed and efficiency? 2. What are the key advantages of using interconnected layers across processing nodes in a CNN?

Frequently Updated Research:

• Stay updated on advancements in parallel processing techniques for neural networks.
• Monitor research on optimizing CNN architectures for specific applications.

Original Abstract Submitted

a parallel convolutional neural network is provided. the cnn is implemented by a plurality of convolutional neural networks each on a respective processing node. each cnn has a plurality of layers. a subset of the layers are interconnected between processing nodes such that activations are fed forward across nodes. the remaining subset is not so interconnected.