Unsupervised changes in core object recognition behavioral performance are accurately predicted by unsupervised neural plasticity in inferior temporal cortex


引用 0|浏览9
Temporal continuity of object identity is a natural feature of visual input statistics, and it is potentially exploited -in an unsupervised manner -by the ventral visual stream to build and re-shape the neural representation in inferior temporal (IT) cortex and ITdependent core object recognition behavior. Prior psychophysical studies in humans and electrophysiological studies in monkey IT are individually supportive of this hypothesis. However, due to differences in tasks and experience manipulations, it is not yet known if the reported plasticity of individual IT neurons and the reported human behavioral changes are quantitatively consistent. Here we tested that consistency by building an unsupervised plasticity model that captures the previously-reported IT neural plasticity and combined that model with a previously established IT-to-recognition-behavior linking model. We compared the predictions of the overall model with the results of three new human behavioral experiments: in each we delivered a different type of unsupervised temporal contiguity experience and longitudinally measured its effect on performance of targeted object discrimination tasks. We found that, without any parameter tuning, the overall model accurately predicted the mean direction, magnitude and time course of performance changes in all three of these experiments. We also found a previously unreported dependency of the observed human performance change on the initial difficulty of the targeted object discrimination task, which was also largely predicted by the overall model. This result demonstrates the interlocking consistency of a range of prior neural and behavioral work, and thus adds support to the hypothesis that tolerant core object recognition in human and non-human primates is instructed -at least in part -by naturally occurring unsupervised temporal contiguity experience.
AI 理解论文