I use machine learning, psychophysics and fMRI to understand how visual processing in healthy human brains allows us to understand the surfaces, materials, shapes, and objects present in an image. I’m interested in neural representation, learning, and (ultimately) consciousness. I am also fascinated by the history and philosophy of our attempts to understand perception..
How brains and machines learn material properties
Unsupervised machine learning may hold the key to fully explicit, image-computable models of how we learn meaningful perceptual dimensions from natural data. My current research programme focuses on whether unsupervised learning goals such as prediction and compression might explain some of the idiosyncrasies in human perception, particularly in mid-level domains such as shape and material.
currently in prep.
Object recognition in humans and deep neural networks
In the past decade, artificial neural networks have finally rivalled human performance in difficult tasks such as recognising objects in cluttered natural images. In Cambridge with Nikolaus Kriegeskorte, my postdoc work focussed on whether these 'engineering solutions' might solve the problems of vision in similar ways to human brains, and how we can create more biologically realistic models.
Storrs, K. R. & Kriegeskorte, N. (in press). Deep learning for cognitive neuroscience. In M. Gazzaniga (Ed.), The Cognitive Neurosciences (6th Edition). Boston: MIT Press.
Jóźwik, K., Kriegeskorte, N., Storrs, K. R., & Mur, M. (2017). Deep convolutional neural networks outperform feature-based but not categorical models in explaining object similarity judgements. Frontiers in Psychology, 8.
In my PhD research, I used psycho-physics and population code modelling to explore what visual aftereffects can tell us about the neural encoding of spatial patterns throughout the visual hierarchy, from simple visual stimuli like tilted lines, through inter-mediate shapes, to complex stimuli like faces.
In particular, many of my studies addressed the question of whether faces are encoded in the brain as individual 'exemplars,' or in terms of how they differ from a special 'norm' such as the average face.
Storrs, K. R. (2015). Are high-level aftereffects perceptual? Frontiers in Psychology, 6(157), 1–4.
Storrs, K. R. & Arnold, D. H. (2017). Shape adaptation exaggerates shape differences. Journal of Experimental Psychology: Human Perception and Performance, 43(1), 181-191.
Storrs, K. R. & Arnold, D. H. (2015). Face aftereffects involve local repulsion, not renormalization. Journal of Vision, 15(8), 1–18.
Storrs, K. R. & Arnold, D. H. (2012). Not all face aftereffects are equal. Vision Research, 64, 7–16.