BI 075 Jim DiCarlo: Reverse Engineering Vision
06/24/20 • 76 min
Jim and I discuss his reverse engineering approach to visual intelligence, using deep models optimized to perform object recognition tasks. We talk about the history of his work developing models to match the neural activity in the ventral visual stream, how deep learning connects with those models, and some of his recent work: adding recurrence to the models to account for more difficult object recognition, using unsupervised learning to account for plasticity in the visual stream, and controlling neural activity by creating specific images for subjects to view.
Notes:
- The DiCarlo Lab at MIT.
- Related papers:
- Large-Scale, High-Resolution Comparison of the Core Visual Object Recognition Behavior of Humans, Monkeys, and State-of-the-Art Deep Artificial Neural Networks.
- Fast recurrent processing via ventral prefrontal cortex is needed by the primate ventral stream for robust core visual object recognition.
- Unsupervised changes in core object recognition behavioral performance are accurately predicted by unsupervised neural plasticity in inferior temporal cortex.
- Neural population control via deep image synthesis.
Jim and I discuss his reverse engineering approach to visual intelligence, using deep models optimized to perform object recognition tasks. We talk about the history of his work developing models to match the neural activity in the ventral visual stream, how deep learning connects with those models, and some of his recent work: adding recurrence to the models to account for more difficult object recognition, using unsupervised learning to account for plasticity in the visual stream, and controlling neural activity by creating specific images for subjects to view.
Notes:
- The DiCarlo Lab at MIT.
- Related papers:
- Large-Scale, High-Resolution Comparison of the Core Visual Object Recognition Behavior of Humans, Monkeys, and State-of-the-Art Deep Artificial Neural Networks.
- Fast recurrent processing via ventral prefrontal cortex is needed by the primate ventral stream for robust core visual object recognition.
- Unsupervised changes in core object recognition behavioral performance are accurately predicted by unsupervised neural plasticity in inferior temporal cortex.
- Neural population control via deep image synthesis.
Previous Episode
BI 074 Ginger Campbell: Are You Sure?
Ginger and I discuss her book Are You Sure? The Unconscious Origins of Certainty, which summarizes Richard Burton's work exploring the experience and phenomenal origin of feeling confident, and how the vast majority of our brain processing occurs outside our conscious awareness.
Next Episode
BI 076 Olaf Sporns: Network Neuroscience
Olaf and I discuss the explosion of network neuroscience, which uses network science tools to map the structure (connectome) and activity of the brain at various spatial and temporal scales. We talk about the possibility of bridging physical and functional connectivity via communication dynamics, and about the relation between network science and artificial neural networks and plenty more.
Notes:
If you like this episode you’ll love
Episode Comments
Featured in these lists
Generate a badge
Get a badge for your website that links back to this episode
<a href="https://goodpods.com/podcasts/brain-inspired-206251/bi-075-jim-dicarlo-reverse-engineering-vision-21442021"> <img src="https://storage.googleapis.com/goodpods-images-bucket/badges/generic-badge-1.svg" alt="listen to bi 075 jim dicarlo: reverse engineering vision on goodpods" style="width: 225px" /> </a>
Copy