18514069. HARDWARE IP OPTIMIZED CONVOLUTIONAL NEURAL NETWORK simplified abstract (Intel Corporation)

HARDWARE IP OPTIMIZED CONVOLUTIONAL NEURAL NETWORK

Organization Name

Intel Corporation

Inventor(s)

Amit Bleiweiss of Yad Binyamin (IL)

Itamar Ben-ari of Givat HaShlosha (IL)

Michael Behar of Zichron Yaakov (IL)

Guy Jacob of Netanya (IL)

Gal Leibovich of Kiryat Yam (IL)

Jacob Subag of Kiryat Haim (IL)

Lev Faivishevsky of Kfar Saba (IL)

Yaniv Fais of Tel Aviv (IL)

Tomer Schwartz of Even Yehuda (IL)

HARDWARE IP OPTIMIZED CONVOLUTIONAL NEURAL NETWORK - A simplified explanation of the abstract

This abstract first appeared for US patent application 18514069 titled 'HARDWARE IP OPTIMIZED CONVOLUTIONAL NEURAL NETWORK

Simplified Explanation

The apparatus described in the abstract is designed to receive a trained neural network model and convert it into an optimized model suitable for a specific execution platform. This process involves adjusting parameters to ensure optimal performance on the chosen platform.

• The apparatus includes at least one execution platform.
• The logic, which includes hardware logic, receives a trained neural network model in a model optimizer.
• The logic then converts the trained neural network model into an optimized model with parameters tailored to the execution platform.

Potential Applications

This technology could be applied in various fields such as:

• Artificial intelligence
• Machine learning
• Data analysis

Problems Solved

This innovation addresses the following issues:

• Optimizing neural network models for specific execution platforms
• Improving performance and efficiency of neural networks
• Streamlining the process of converting trained models for deployment

Benefits

The benefits of this technology include:

• Enhanced performance of neural network models
• Increased efficiency in model deployment
• Simplified optimization process for different platforms

Potential Commercial Applications

A potential commercial application for this technology could be in:

• Developing custom neural network models for specific hardware platforms

Possible Prior Art

One possible prior art for this technology could be:

• Existing methods for optimizing neural network models for different hardware architectures

How does this technology compare to existing methods for optimizing neural network models for specific execution platforms?

This article does not provide a direct comparison with existing methods for optimizing neural network models. It would be beneficial to understand the specific advantages and disadvantages of this new approach compared to traditional techniques.

What impact could this technology have on the efficiency and performance of neural network models in real-world applications?

While the benefits of the technology are outlined, the article does not delve into the potential real-world impact on efficiency and performance. Understanding how this innovation could improve outcomes in practical scenarios would provide valuable insights for potential users.

Original Abstract Submitted

In an example, an apparatus comprises at least one execution platform; and logic, at least partially including hardware logic, to receive a trained neural network model in a model optimizer and convert the trained neural network model to an optimized model comprising parameters that are fit to the at least one execution platform. Other embodiments are also disclosed and claimed.