Description
Neural Network Libraries, developed by Sony, is an open source software aimed at making the research, development, and implementation of neural networks more efficient. It allows users to define computation graphs intuitively with minimal code, supporting both static and dynamic computation graph paradigms. This flexibility enables runtime network construction, which is crucial for various applications.
Read More
The library is designed with portability in mind, running on both Linux and Windows platforms. Most of its code is written in C++14, allowing deployment onto embedded devices. This makes it suitable for a wide range of applications, from mobile devices to complex systems.
Neural Network Libraries also feature a function abstraction and code template generator, making it easier for developers to add new functions with less coding effort. Additionally, it supports multi-target device acceleration through plugins, with CUDA implemented as a plugin extension.
Sony has utilized Neural Network Libraries in various products, such as the Real Estate Price Estimate Engine, which analyzes massive data to estimate prices in real estate transactions. It is also used in the Xperia Ear for gesture sensitivity functions and in the aibo robot for image recognition tasks like user identification and face tracking.
Overall, Neural Network Libraries provide a robust platform for developing neural network applications, offering flexibility, portability, and ease of use for developers.
- CategoryMachine Learning Platforms
- Pricing ModelOpen source
- DeploymentCloud, Self-hosted
- PlatformsLinux, Windows
Neural Network Libraries's Core Features
Open source software for neural networks
Supports static and dynamic computation graphs
Portable across Linux and Windows
C++14 core API for embedded devices
Function abstraction and code template generator
Multi-target device acceleration via plugins
CUDA implemented as a plugin extension
Used in Sony's real estate and mobile products
How to use Neural Network Libraries?
Define: Create computation graphs with minimal code
Deploy: Use C++14 API for embedded devices
Extend: Add new functions using templates
Accelerate: Implement device acceleration with plugins
Neural Network Libraries's Use Cases
- Real Estate Analysis
- Gesture Sensitivity
- Image Recognition
- Embedded Device Deployment
- Dynamic Network Construction






