Date
2022
details
CEA-List
Developing a demonstrator to showcase the CEA’s latest AI-robustness assessment technology.
A team of researchers at CEA-List, a research lab part of France’s public technology and energy research organization, has been developing a new software named PyRAT. Designed to prove the reliability of neural networks’ response in given configurations, PyRAT can help assess the trustworthiness of AI-driven systems in critical applications ranging from aircraft collision avoidance to medical diagnostics.
The laboratory was eager to demonstrate PyRAT to a wide audience, including researchers, scientists, industrial partners or prospects in a way both technical and educational, through a demonstrator to be presented in the CEA’s showroom in Paris-Saclay. We worked with the laboratory on how to best represent and explain the highly complex and opaque processes at work in large neural networks.
The result is a visually-rich software interface uncovering the logic behind PyRAT and the various mathematical operations at play, through which users can experiment with different analysis parameters. The interface is on display inside the CEA’s showrooms where it can be controlled through a dedicated hardware controller that we designed and developed, embodying by itself the laboratory’s scientific breakthrough.
A team of researchers at CEA-List, a research lab part of France’s public technology and energy research organization, has been developing a new software named PyRAT. Designed to prove the reliability of neural networks’ response in given configurations, PyRAT can help assess the trustworthiness of AI-driven systems in critical applications ranging from aircraft collision avoidance to medical diagnostics.
The laboratory was eager to demonstrate PyRAT to a wide audience, including researchers, scientists, industrial partners or prospects in a way both technical and educational, through a demonstrator to be presented in the CEA’s showroom in Paris-Saclay. We worked with the laboratory on how to best represent and explain the highly complex and opaque processes at work in large neural networks.
The result is a visually-rich software interface uncovering the logic behind PyRAT and the various mathematical operations at play, through which users can experiment with different analysis parameters. The interface is on display inside the CEA’s showrooms where it can be controlled through a dedicated hardware controller that we designed and developed, embodying by itself the laboratory’s scientific breakthrough.
A team of researchers at CEA-List, a research lab part of France’s public technology and energy research organization, has been developing a new software named PyRAT. Designed to prove the reliability of neural networks’ response in given configurations, PyRAT can help assess the trustworthiness of AI-driven systems in critical applications ranging from aircraft collision avoidance to medical diagnostics.
The laboratory was eager to demonstrate PyRAT to a wide audience, including researchers, scientists, industrial partners or prospects in a way both technical and educational, through a demonstrator to be presented in the CEA’s showroom in Paris-Saclay. We worked with the laboratory on how to best represent and explain the highly complex and opaque processes at work in large neural networks.
The result is a visually-rich software interface uncovering the logic behind PyRAT and the various mathematical operations at play, through which users can experiment with different analysis parameters. The interface is on display inside the CEA’s showrooms where it can be controlled through a dedicated hardware controller that we designed and developed, embodying by itself the laboratory’s scientific breakthrough.
A team of researchers at CEA-List, a research lab part of France’s public technology and energy research organization, has been developing a new software named PyRAT. Designed to prove the reliability of neural networks’ response in given configurations, PyRAT can help assess the trustworthiness of AI-driven systems in critical applications ranging from aircraft collision avoidance to medical diagnostics.
The laboratory was eager to demonstrate PyRAT to a wide audience, including researchers, scientists, industrial partners or prospects in a way both technical and educational, through a demonstrator to be presented in the CEA’s showroom in Paris-Saclay. We worked with the laboratory on how to best represent and explain the highly complex and opaque processes at work in large neural networks.
The result is a visually-rich software interface uncovering the logic behind PyRAT and the various mathematical operations at play, through which users can experiment with different analysis parameters. The interface is on display inside the CEA’s showrooms where it can be controlled through a dedicated hardware controller that we designed and developed, embodying by itself the laboratory’s scientific breakthrough.
