Direct Fit to Nature: An Evolutionary Perspective on Biological and Artificial Neural Networks

Abstract

Evolution is a blind fitting process by which organisms become adapted to their environment. Does the brain use similar brute-force fitting processes to learn how to perceive and act upon the world? Recent advances in artificial neural networks have exposed the power of optimizing millions of synaptic weights over millions of observations to operate robustly in real-world contexts. These models do not learn simple, human-interpretable rules or representations of the world; rather, they use local computations to interpolate over task-relevant manifolds in a high-dimensional parameter space. Counterintuitively, similar to evolutionary processes, over-parameterized models can be simple and parsimonious, as they provide a versatile, robust solution for learning a diverse set of functions. This new family of direct-fit models present a radical challenge to many of the theoretical assumptions in psychology and neuroscience. At the same time, this shift in perspective establishes unexpected links with developmental and ecological psychology.

Convolutional Neural Networks as a Model of the Visual System: Past, Present, and Future

“…The shift towards models that actually do something has forced a reshaping of the questions around the study of the visual system. Neuroscientists are adapting to this new style of explanation and the different expectations that come with it [121].…”

Contrasting action and posture coding with hierarchical deep neural network models of proprioception

“…We iteratively constrain the solution space by choosing joint angles in vicinity of the previous configuration in order to eliminate redundancy. To cover a large 3D workspace we placed the characters in multiple horizontal (26) and vertical (18) planes and calculated corresponding joint-angle trajectories (starting points are shown in Figure 2D). From this set, we selected a subset of two hundred thousand examples with smooth, non-jerky muscle length changes, while making sure that the set is balanced in terms of the number of examples per class (see Methods).…”

“…networks learn image representations that closely resemble tuning properties of single neurons in the ventral pathway of primates and elucidate likely transformations along the ventral stream (17,(19)(20)(21)(22)(23)(24). This goal-driven modelling approach (17,(25)(26)(27) has since successfully been applied to other sensory modalities such as touch (28,29), thermosensation (30) and audition (31).…”

Decoding subjective emotional arousal from EEG during an immersive Virtual Reality experience

“…VR as a step towards a real-world neuroscience More naturalistic experimental stimulation, for example using immersive VR, allows to test the brain under conditions it was optimized for and thereby improve the discovery of neural features and dynamics (Gibson, 1979;Hasson et al, 2020). Findings from naturalistic studies can test the real-world relevance of results obtained in highly controlled, abstract laboratory settings (Matusz et al, 2019;Shamay-Tsoory & Mendelsohn, 2019).…”