Skip to content
Drix10 Blog

Parametric Neural Control Differentiates Top Models

, 2 items in AI Organizations and Media, 2 min read

In this digest (2 items)

Parametric neural control reveals hidden differences among vision models that predict neural responses equally well. The method synthesizes images to drive specific firing rates and compares model predictions to actual neural modulation.

Key points

  • Hundreds of distinct models achieve nearly identical prediction scores on natural images.

  • ResNet50 encoding axis achieved r = 0.87 on calibration, r = 0.95 on held-out images, but r = 0.28 on axis-aligned accentuation images.

Sources

Axel Krieger talk Autonomous Robotic Surgery – 10:30-11:45

Axel Krieger will present “Autonomous Robotic Surgery: From Teleoperation to Learning Systems” at 10:30 am–11:45 am. The seminar is part of the GRASP on Robotics series and is in-person at Wu and Chen Auditorium. It covers the Smart Tissue Autonomous Robot and the Surgical Robot Transformer (SRT-H).

Key points

  • Talk time: 10:30 am-11:45 am

  • Venue: Wu and Chen Auditorium, in-person only

Sources