Representation space

Continual learning (CL), the capability to acquire knowledge and skills over time without forgetting old ones, is fundamental to how animals survive in a non-stationary world. This ability has proven difficult to realize in artificial neural networks (ANNs) [1], which suffer from catastrophic forgetting of previously learned information. To understand this contrast, we sought to understand what aspects of a neural system and the tasks it learns are key to its CL performance. We developed a theory of deep, nonlinear feedforward networks sequentially learning multiple tasks and derived exact exp

Biological visual systems learn complex representations of the world that support a wide range of cognitive behaviors without using a large number of labelled examples. The efficient coding hypothesis suggests that this is accomplished by adapting the sensory representation to the statistics of the input signal, i.e. in a way that facilitates redundancy reduction. Visual signals have several clear sources of redundancy. They evolve slowly in time, since temporally adjacent inputs typically correspond to different views of the same scene, which in turn are usually more similar than views of dis

…of weeks, a phenomenon often called ‘representational drift’. For example, in piriform cortex, which is commonly considered primary olfactory cortex, the cells responsive to a given odor are completely uncorrelated with those activated by the same odor two weeks later (Schoonover et al., 2021). This would seem to imply that piriform cortex, like other brain regions whose activity appears to drift, is useless for the retrieval of associative memories learned several weeks prior. Howeve