A neural network account of memory replay and knowledge consolidation

Demystifying unsupervised learning: how it helps and hurts

Humans and machines rarely have access to explicit external feedback or supervision, yet manage to learn. Most modern machine learning systems succeed because they benefit from unsupervised data. Humans are also expected to benefit and yet, mysteriously, empirical results are mixed. Does unsupervised learning help humans or not? Here, we argue that the mixed results are not conflicting answers to this question, but reflect that humans self-reinforce their predictions in the absence of supervision, which can help or hurt depending on whether predictions and task align. We use this framework to synthesize empirical results across various domains to clarify when unsupervised learning will help or hurt. This provides new insights into the fundamentals of learning with implications for instruction and lifelong learning.

Differences in the neural correlates of schizophrenia with positive and negative formal thought disorder in patients with schizophrenia in the ENIGMA dataset

Formal thought disorder (FTD) is a clinical key factor in schizophrenia, but the neurobiological underpinnings remain unclear. In particular, the relationship between FTD symptom dimensions and patterns of regional brain volume loss in schizophrenia remains to be established in large cohorts. Even less is known about the cellular basis of FTD. Our study addresses these major obstacles by enrolling a large multi-site cohort acquired by the ENIGMA Schizophrenia Working Group (752 schizophrenia patients and 1256 controls), to unravel the neuroanatomy of FTD in schizophrenia and using virtual histology tools on implicated brain regions to investigate the cellular basis. Based on the findings of previous clinical and neuroimaging studies, we decided to separately explore positive, negative and total formal thought disorder. We used virtual histology tools to relate brain structural changes associated with FTD to cellular distributions in cortical regions. We identified distinct neural networks positive and negative FTD. Both networks encompassed fronto-occipito-amygdalar brain regions, but positive and negative FTD demonstrated a dissociation: negative FTD showed a relative sparing of orbitofrontal cortical thickness, while positive FTD also affected lateral temporal cortices. Virtual histology identified distinct transcriptomic fingerprints associated for both symptom dimensions. Negative FTD was linked to neuronal and astrocyte fingerprints, while positive FTD also showed associations with microglial cell types. These results provide an important step towards linking FTD to brain structural changes and their cellular underpinnings, providing an avenue for a better mechanistic understanding of this syndrome.

Manipulated overlapping reactivation of multiple memories promotes explicit gist abstraction

Sleep is considered to promote gist abstraction on the basis of spontaneous memory reactivation. As speculated in the theory of 'information overlap to abstract (iOtA)', 'overlap' between reactivated memories, beyond reactivation, is crucial to gist abstraction. Yet so far, empirical research has not tested this theory by manipulating the factor of 'overlap'. In the current study, 'overlap' itself was manipulated by targeted memory reactivation (TMR), through simultaneously reactivating multiple memories that either contain or do not contain spatially overlapped gist information, to investigate the effect of overlapping reactivation on gist abstraction. This study had a factorial design of 2 factors with 2 levels respectively (spatial overlap/no spatial overlap, TMR/no-TMR). Accordingly, 82 healthy college students (aged 19 ∼ 25, 57 females) were randomized into four groups. After learning 16 pictures, paired with 4 auditory cues (4 pictures - 1 cue) according to the grouping, participants were given a 90-minute nap opportunity. Then TMR cueing was conducted during N2 and slow wave sleep of the nap. Performance in memory task was used to measure gist abstraction. The results showed a significant main effect of TMR on both implicit and explicit gist abstraction, and a marginally significant interaction effect on explicit gist abstraction. Further analyses showed that explicit gist abstraction in the spatial overlap & TMR group was significantly better than in the control group. Moreover, explicit gist abstraction was positively correlated with spindle density. The current study thus indicates that TMR facilitates gist abstraction, and explicit gist abstraction may benefit more from overlapping reactivation.

Shared structure facilitates working memory of multiple sequences

Daily experiences often involve the processing of multiple sequences, yet storing them challenges the limited capacity of working memory (WM). To achieve efficient memory storage, relational structures shared by sequences would be leveraged to reorganize and compress information. Here, participants memorized a sequence of items with different colors and spatial locations and later reproduced the full color and location sequences one after another. Crucially, we manipulated the consistency between location and color sequence trajectories. First, sequences with consistent trajectories demonstrate improved memory performance and a trajectory correlation between reproduced color and location sequences. Second, sequences with consistent trajectories show neural reactivation of common trajectories, and display spontaneous replay of color sequences when recalling locations. Finally, neural reactivation correlates with WM behavior. Our findings suggest that a shared common structure is leveraged for the storage of multiple sequences through compressed encoding and neural replay, together facilitating efficient information organization in WM.

Distinguishable neural circuit mechanisms associated with the clinical efficacy of rTMS in aMCI patients

Repetitive transcranial magnetic stimulation is used in early-stage Alzheimer's disease to slow progression, but heterogeneity in response results in different treatment outcomes. The mechanisms underlying this heterogeneity are unclear. This study used resting-state neuroimaging to investigate the variability in episodic memory improvement from angular gyrus repetitive transcranial magnetic stimulation and tracked the neural circuits involved. Thirty-four amnestic mild cognitive impairment patients underwent angular gyrus repetitive transcranial magnetic stimulation (4 weeks, 20 Hz, 100% resting motor threshold) and were divided into high-response and low-response groups based on minimal clinically important differences in auditory verbal learning test scores. Baseline and pre/post-treatment neural circuit activities were compared. Results indicated that the orbital middle frontal gyrus in the orbitofrontal cortex network and the precuneus in the default mode network had higher local activity in the low-response group. After treatment, changes in local and remote connectivity within brain regions of the orbitofrontal cortex, default mode network, visual network, and sensorimotor network showed opposite trends and were related to treatment effects. This suggests that the activity states of brain regions within the orbitofrontal cortex and default mode network could serve as imaging markers for early cognitive compensation in amnestic mild cognitive impairment patients and predict the aftereffects of repetitive transcranial magnetic stimulation response.

A generative model of memory construction and consolidation

Episodic memories are (re)constructed, share neural substrates with imagination, combine unique features with schema-based predictions and show schema-based distortions that increase with consolidation. Here we present a computational model in which hippocampal replay (from an autoassociative network) trains generative models (variational autoencoders) to (re)create sensory experiences from latent variable representations in entorhinal, medial prefrontal and anterolateral temporal cortices via the hippocampal formation. Simulations show effects of memory age and hippocampal lesions in agreement with previous models, but also provide mechanisms for semantic memory, imagination, episodic future thinking, relational inference and schema-based distortions including boundary extension. The model explains how unique sensory and predictable conceptual elements of memories are stored and reconstructed by efficiently combining both hippocampal and neocortical systems, optimizing the use of limited hippocampal storage for new and unusual information. Overall, we believe hippocampal replay training generative models provides a comprehensive account of memory construction, imagination and consolidation.

Generative replay underlies compositional inference in the hippocampal-prefrontal circuit

Human reasoning depends on reusing pieces of information by putting them together in new ways. However, very little is known about how compositional computation is implemented in the brain. Here, we ask participants to solve a series of problems that each require constructing a whole from a set of elements. With fMRI, we find that representations of novel constructed objects in the frontal cortex and hippocampus are relational and compositional. With MEG, we find that replay assembles elements into compounds, with each replay sequence constituting a hypothesis about a possible configuration of elements. The content of sequences evolves as participants solve each puzzle, progressing from predictable to uncertain elements and gradually converging on the correct configuration. Together, these results suggest a computational bridge between apparently distinct functions of hippocampal-prefrontal circuitry and a role for generative replay in compositional inference and hypothesis testing.

Signatures of cross-modal alignment in children's early concepts

Whether supervised or unsupervised, human and machine learning is usually characterized as event-based. However, learning may also proceed by systems alignment in which mappings are inferred between entire systems, such as visual and linguistic systems. Systems alignment is possible because items that share similar visual contexts, such as a car and a truck, will also tend to share similar linguistic contexts. Because of the mirrored similarity relationships across systems, the visual and linguistic systems can be aligned at some later time absent either input. In a series of simulation studies, we considered whether children's early concepts support systems alignment. We found that children's early concepts are close to optimal for inferring novel concepts through systems alignment, enabling agents to correctly infer more than 85% of visual-word mappings absent supervision. One possible explanation for why children's early concepts support systems alignment is that they are distinguished structurally by their dense semantic neighborhoods. Artificial agents using these structural features to select concepts proved highly effective, both in environments mirroring children's conceptual world and those that exclude the concepts that children commonly acquire. For children, systems alignment and event-based learning likely complement one another. Likewise, artificial systems can benefit from incorporating these developmental principles.

Neural Correlates of Positive and Negative Formal Thought Disorder in Individuals with Schizophrenia: An ENIGMA Schizophrenia Working Group Study

Formal thought disorder (FTD) is a key clinical factor in schizophrenia, but the neurobiological underpinnings remain unclear. In particular, relationship between FTD symptom dimensions and patterns of regional brain volume deficiencies in schizophrenia remain to be established in large cohorts. Even less is known about the cellular basis of FTD. Our study addresses these major obstacles based on a large multi-site cohort through the ENIGMA Schizophrenia Working Group (752 individuals with schizophrenia and 1256 controls), to unravel the neuroanatomy of positive, negative and total FTD in schizophrenia and their cellular bases. We used virtual histology tools to relate brain structural changes associated with FTD to cellular distributions in cortical regions. We identified distinct neural networks for positive and negative FTD. Both networks encompassed fronto-occipito-amygdalar brain regions, but negative FTD showed a relative sparing of orbitofrontal cortical thickness, while positive FTD also affected lateral temporal cortices. Virtual histology identified distinct transcriptomic fingerprints associated for both symptom dimensions. Negative FTD was linked to neuronal and astrocyte fingerprints, while positive FTD was also linked to microglial cell types. These findings relate different dimensions of FTD to distinct brain structural changes and their cellular underpinnings, improve our mechanistic understanding of these key psychotic symptoms.

Neural Correlates of Positive and Negative Formal Thought Disorder in Individuals with Schizophrenia: An ENIGMA Schizophrenia Working Group Study

Formal thought disorder (FTD) is a key clinical factor in schizophrenia, but the neurobiological underpinnings remain unclear. In particular, relationship between FTD symptom dimensions and patterns of regional brain volume deficiencies in schizophrenia remain to be established in large cohorts. Even less is known about the cellular basis of FTD. Our study addresses these major obstacles based on a large multi-site cohort through the ENIGMA Schizophrenia Working Group (752 individuals with schizophrenia and 1256 controls), to unravel the neuroanatomy of positive, negative and total FTD in schizophrenia and their cellular bases. We used virtual histology tools to relate brain structural changes associated with FTD to cellular distributions in cortical regions. We identified distinct neural networks for positive and negative FTD. Both networks encompassed fronto-occipito-amygdalar brain regions, but negative FTD showed a relative sparing of orbitofrontal cortical thickness, while positive FTD also affected lateral temporal cortices. Virtual histology identified distinct transcriptomic fingerprints associated for both symptom dimensions. Negative FTD was linked to neuronal and astrocyte fingerprints, while positive FTD was also linked to microglial cell types. These findings relate different dimensions of FTD to distinct brain structural changes and their cellular underpinnings, improve our mechanistic understanding of these key psychotic symptoms.

How our understanding of memory replay evolves

Memory reactivations and replay, widely reported in the hippocampus and cortex across species, have been implicated in memory consolidation, planning, and spatial and skill learning. Technological advances in electrophysiology, calcium imaging, and human neuroimaging techniques have enabled neuroscientists to measure large-scale neural activity with increasing spatiotemporal resolution and have provided opportunities for developing robust analytic methods to identify memory replay. In this article, we first review a large body of historically important and representative memory replay studies from the animal and human literature. We then discuss our current understanding of memory replay functions in learning, planning, and memory consolidation and further discuss the progress in computational modeling that has contributed to these improvements. Next, we review past and present analytic methods for replay analyses and discuss their limitations and challenges. Finally, looking ahead, we discuss some promising analytic methods for detecting nonstereotypical, behaviorally nondecodable structures from large-scale neural recordings. We argue that seamless integration of multisite recordings, real-time replay decoding, and closed-loop manipulation experiments will be essential for delineating the role of memory replay in a wide range of cognitive and motor functions.