Integration of New Information in Memory: New Insights from a Complementary Learning Systems Perspective

By James L. McClelland, Bruce L. McNaughton, Andrew Kyle Lampinen. 2020

113 citations recorded by the source corpus.

According to complementary learning systems theory, integrating new memories into the neocortex of the brain without interfering with what is already known depends on a gradual learning process, interleaving new items with previously learned items. However, empirical studies show that information consistent with prior knowledge can be integrated very quickly. We use artificial neural networks with properties like those we attribute to the neocortex to develop a theoretical understanding of the role of consistency with prior knowledge in putatively neocortex-like learning systems, providing new insights into when integration will be fast or slow and how integration might be made more efficient when the items to be learned are hierarchically structured. The work relies on deep linear networks that capture the qualitative aspects of the learning dynamics of the more complex non-linear networks used in previous work. The time course of learning in these networks can be linked to the hierarchical structure in the training data, captured mathematically as a set of dimensions that correspond to the branches in the hierarchy. In this context, a new item to be learned can be characterized as having aspects that project onto previously known dimensions, and others that require adding a new branch/dimension. The projection onto the known dimensions can be learned rapidly without interleaving, but learning the new dimension requires gradual interleaved learning. When a new item only overlaps with items within one branch of a hierarchy, interleaving can focus on the previously-known items within this branch, resulting in faster integration with less inter-leaving overall. The discussion considers how the brain might exploit these facts to make learning more efficient and highlights predictions about what aspects of new information might be hard or easy to learn.

Subfields: Cognitive Neuroscience, Developmental and Educational Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Cognitive Neuroscience

In Developmental and Educational Psychology

Citations in the corpus

Replay in Deep Learning: Current Approaches and Missing Biological Elements (2021, neutral): “In the mammalian brain, one mechanism used to combat forgetting and facilitate consolidation is replay: the reactivation of past neural activation patterns 1 (McClelland, McNaughton, & O’Reilly, 1995; Kumaran, Hassabis, & McClelland, 2016; McClelland, McNaughton, & Lampinen, 2020).”
Towards mental time travel: a hierarchical memory for reinforcement learning agents (2021, background): “For example, could memory summaries be used to locate similar or contrasting memories that should be interleaved together [38] to improve continual learning?”
Biological underpinnings for lifelong learning machines (2022, background): “…work in deep learning has addressed the issue of scalability by showing that to avoid forgetting, it can be sufficient to only replay a small subset 54 , to just replay old memories that are similar to the new learning 203 , or to replay abstract, high-level representations of past experiences 54 .”
Class Similarity Weighted Knowledge Distillation for Continual Semantic Segmentation (2022, background): “A recent neuroscience study investigates how new knowledge is integrated into a neocortex-like network [22].”
Selective Replay Enhances Learning in Online Continual Analogical Reasoning (2021, background): “Although replay selection policies have been explored for supervised classi-ﬁcation [5, 6, 15, 57, 83], they have not yielded signiﬁcant improvements, especially on large-scale datasets [34, 83].”
Extending Machine Language Models toward Human-Level Language Understanding (2019, background): “In CLS (58), the neocortex learns arbitrary new information gradually through interleaved presentations of new and familiar items, weaving it into the fabric of knowledge and avoiding interference with existing knowledge (60).”
What shapes feature representations? Exploring datasets, architectures, and training (2020, background): “” features from CNNs trained on naturalistic datasets. However, catastrophic interference between new tasks and prior tasks shows that at least some features are suppressed or altered during training [28, 21, 46, 33, 27]. Thus, a tension exists in the literature about how features change over learni”