IDiyas

For the Inventor, By the Inventor

Urayasu, Japan

Taro Sekiyama

USPTO Granted Patents = 10 

Average Co-Inventor Count = 3.8

ph-index = 2

Forward Citations = 14(Granted Patents)


Location History:

  • Urayasu, JP (2019 - 2022)
  • Chiba, JP (2024)

Company Filing History:

goldMedal10 out of 164,108 
 
International Business Machines Corporation
 patents

Years Active: 2019-2024

Loading Chart...
10 patents (USPTO):
1. 11880762 - Choosing execution mode of a neural network based on total memory usage (2024-01-23),
2. 11461637 - Real-time resource usage reduction in artificial neural networks (2022-10-04),
3. 11164079 - Multi-GPU deep learning using CPUs (2021-11-02),
4. 11106970 - Localizing tree-based convolutional neural networks (2021-08-31),
5. 10915810 - Weighted cascading convolutional neural networks (2021-02-09),
6. 10782897 - Memory reduction for neural networks with fixed structures (2020-09-22),
7. 10599978 - Weighted cascading convolutional neural networks (2020-03-24),
8. 10558914 - Real-time resource usage reduction in artificial neural networks (2020-02-11),
9. 10268951 - Real-time resource usage reduction in artificial neural networks (2019-04-23),
10. 10169874 - Surface-based object identification (2019-01-01),

Title: Taro Sekiyama: Innovator in Neural Network Technologies

Introduction

Taro Sekiyama is a prominent inventor based in Urayasu, Japan. He has made significant contributions to the field of artificial intelligence, particularly in the development of neural network technologies. With a total of 10 patents to his name, Sekiyama is recognized for his innovative approaches to optimizing neural network performance.

Latest Patents

Among his latest patents, Sekiyama has developed a method for choosing execution modes of a neural network based on total memory usage. This invention provides a computer-implemented method, a computer program product, and a computer processing system for selecting from multiple Graphics Processing Unit (GPU) execution modes for a Neural Network (NN) that exceeds a certain size. The execution modes include normal memory mode, Out-of-Core (OoC) execution mode, and Unified Memory (UM) mode. The method initiates execution on the NN using the UM mode and measures memory usage across all layers, ultimately selecting the most efficient execution mode based on this data.

Another notable patent focuses on real-time resource usage reduction in artificial neural networks. This innovation captures a generated algorithm during the execution of a neural network iteration. It identifies a candidate algorithm that utilizes less memory than the generated algorithm. Upon this determination, the neural network is updated by replacing the generated algorithm with the more efficient candidate.

Career Highlights

Taro Sekiyama is currently employed at International Business Machines Corporation (IBM), where he continues to push the boundaries of technology in artificial intelligence. His work has garnered attention for its practical applications and potential to enhance computational efficiency in neural networks.

Collaborations

Sekiyama collaborates with talented individuals in his field, including coworkers Yasushi Negishi and Tung Duc Le. Their combined expertise contributes to the innovative projects at IBM, fostering a collaborative environment that encourages groundbreaking advancements.

Conclusion

Taro Sekiyama stands out as a key figure in the realm of neural network technology, with a strong portfolio of patents that reflect his innovative spirit. His contributions are shaping the future of artificial intelligence and optimizing resource usage in complex computational systems.

This text is generated by artificial intelligence and may not be accurate.
Please report any incorrect information to support@idiyas.com
Loading…