Biology subjects

Arun, S.

Publications and source records attributed to Arun, S..

3 recordsLinked to original sources

bioRxiv2019

Do deep neural networks see the way we do?

Deep neural networks have revolutionized computer vision, and their object representations match coarsely with the brain. As a result, it is widely believed that any fine scale differences between deep networks and brains can be fixed with increased training data or minor changes in architecture. But what if there are qualitative differences between brains and deep networks? Do deep networks even see the way we do? To answer this question, we chose a deep neural network optimized for object recognition and asked whether it exhibits well-known perceptual and neural phenomena despite not being explicitly trained to do so. To our surprise, many phenomena were present in the network, including the Thatcher effect, mirror confusion, Webers law, relative size, multiple object normalization and sparse coding along multiple dimensions. However, some perceptual phenomena were notably absent, including processing of 3D shape, patterns on surfaces, occlusion, natural parts and a global advantage. Our results elucidate the computational challenges of vision by showing that learning to recognize objects suffices to produce some perceptual phenomena but not others and reveal the perceptual properties that could be incorporated into deep networks to improve their performance.

neuroscience↗

bioRxiv2019

How the forest interacts with the trees: Multiscale shape integration explains global and local processing

Hierarchical stimuli (such as a circle made of diamonds) have been widely used to study global and local processing. Two classic phenomena have been observed using these stimuli: the global advantage effect (that we identify the circle faster than the diamonds) and the incongruence effect (that we identify the circle faster when both global and local shapes are circles). Understanding them has been difficult because they occur during shape detection, where an unknown categorical judgement is made on an unknown feature representation.\n\nHere we report two essential findings. First, these phenomena are present both in a general same-different task and a visual search task, suggesting that they may be intrinsic properties of the underlying representation. Second, in both tasks, responses were explained using linear models that combined multiscale shape differences and shape distinctiveness. Thus, global and local processing can be understood as properties of a systematic underlying feature representation.

neuroscience↗

bioRxiv2019

Implicit perceptual memory can increase or decrease with ageing

A decline in declarative or explicit memory has been extensively characterized in cognitive ageing and is a hallmark of cognitive impairments. However, whether and how implicit perceptual memory varies with ageing or cognitive impairment is unclear. Here, we compared implicit perceptual memory and explicit memory measures in three groups of subjects: (1) 59 healthy young volunteers (20-30 years); (2) 238 healthy old volunteers (50-90 years) and (3) 21 patients with mild cognitive impairment MCI (50-90 years). To measure explicit memory, subjects were tested on standard recognition and recall tasks. To measure implicit perceptual memory, we used a classic perceptual priming paradigm. Subjects had to report the shape of a visual search pop-out target. Implicit priming was measured as the speedup in response time for targets with the same vs different color/position on consecutive trials.\n\nOur main findings are as follows: (1) Explicit memory was weaker in old compared to young subjects, and in MCI compared to age-matched controls; (2) Surprisingly, implicit perceptual memory did not always decline with age: color priming was smaller in older subjects but position priming was larger; (3) Position priming was less frequent in the MCI group compared to age-matched controls; (4) Implicit and explicit memory measures were uncorrelated in all three groups. Thus, implicit memory can increase or decrease with age or cognitive impairment, but this decline does not covary with explicit memory. We propose that incorporating explicit and implicit measures can yield a richer characterization of memory.

animal behavior and cognition↗