- Indexing
Robustness evaluation method based on maximum safe radius for natural noise
Contributes to improving the quality of AI products by establishing a quantitative index based on AI model robustness.
- In AI service, there may be changes in data trends, for example due to long-term operations. Some AI models, however, cannot respond to targeting data after these changes in trends, and there are cases where instability arises in AI model operations.
- Model robustness evaluations are required to make judgments on whether AI models demonstrate stable operations. Model robustness refers to the degree to which the same output can be obtained whether or not there are fluctuations in the input data. In the case of camera image analysis, there are seasonal and climate variations. Robustness evaluations are important in responding to these fluctuations.
- Robustness evaluation methods based on maximum safe radius for natural noise enable judgments on whether the model demonstrates stable operations through quantitative evaluations.
Applications
- AI product quality management
- Meet societal principles announced by governments of Japan or other countries
- Sharing of AI quality information with the customer
Benchmarks, strengths, and track record
- Enables calculation of certified model robustness and quantitative evaluation of robustness.
Inquiries
Contact the Toshiba Corporate Research & Development Center
Please include the title “Toshiba AI Technology Catalog: Robustness evaluation method based on maximum radius of safety for incidental noise” or the URL in the inquiry text.
Please note that because this technology is currently the subject of R&D activities, immediate responses to inquiries may not be possible.
References:
