Explore the vast range of titles available.

Many facts come with an expiration date, from the name of the President to the basketball team Lebron James plays for. However, most language models (LMs) are trained on snapshots of data collected at a specific moment in time. This can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize. We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum—those trained on specific slices of temporal data, as well as those trained on a wide range of temporal data. To mitigate these problems, we propose a simple technique for jointly modeling text with its timestamp. This improves memorization of seen facts from the training time period, as well as calibration on predictions about unseen facts from future time periods. We also show that models trained with temporal context can be efficiently “refreshed” as new data arrives, without the need for retraining from scratch.

Hippocampal sharp wave ripples (SPW-Rs) have been hypothesized as a mechanism for memory consolidation and action planning. The duration of ripples shows a skewed distribution with a minority of long-duration events. We discovered that long-duration ripples are increased in situations demanding memory in rats. Prolongation of spontaneously occurring ripples by optogenetic stimulation, but not randomly induced ripples, increased memory during maze learning. The neuronal content of randomly induced ripples was similar to short-duration spontaneous ripples and contained little spatial information. The spike content of the optogenetically prolonged ripples was biased by the ongoing, naturally initiated neuronal sequences. Prolonged ripples recruited new neurons that represented either arm of the maze. Long-duration hippocampal SPW-Rs replaying large parts of planned routes are critical for memory.

Memorization is necessary for various fields, such as language learning in the field of education. While memorization learning is often tedious and demotivating due to requiring conscious effort, few support approaches improve memorization unconsciously with low conscious effort. In this study, we propose a method, Mindless Memorization Booster, which improves users’ memorization unconsciously by visual stimuli of modulating the visual interface. This method is based on previous findings that the modulation of perceptual stimuli arouses attention/concentration. When the user looks at the memorization target, the proposed method presents a change in visual interface (e.g., changes in memorization target size, background color, and visual icon movement) to cause a psychological phenomenon of affecting the user’s attention and concentration, aiming at enhancing the memorization unconsciously. A prototype system of the proposed method was implemented for an English vocabulary memorization learning application. The evaluation results showed that the user’s memorization result was affected by the proposed method, and the speed of recall (i.e., outputs of the memorization word from the brain) increased by about 1 s per one memorization word without causing a negative affection on the number of correct answers for memorization. This result indicated the feasibility of the proposed method for memorization learning support. Our findings are helpful for designing visual information interfaces that consider the phenomena affecting the user’s memorization and promote memorization learning unconsciously.

Spaced repetition is a technique for efficient memorization which uses repeated review of content following a schedule determined by a spaced repetition algorithm to improve long-term retention. However, current spaced repetition algorithms are simple rule-based heuristics with a few hard-coded parameters. Here, we introduce a flexible representation of spaced repetition using the framework of marked temporal point processes and then address the design of spaced repetition algorithms with provable guarantees as an optimal control problem for stochastic differential equations with jumps. For two well-known human memory models, we show that, if the learner aims to maximize recall probability of the content to be learned subject to a cost on the reviewing frequency, the optimal reviewing schedule is given by the recall probability itself. As a result, we can then develop a simple, scalable online spaced repetition algorithm, MEMORIZE, to determine the optimal reviewing times. We perform a large-scale natural experiment using data from Duolingo, a popular language-learning online platform, and show that learners who follow a reviewing schedule determined by our algorithm memorize more effectively than learners who follow alternative schedules determined by several heuristics.

The influential notion that the hippocampus supports associative memory by interacting with functionally distinct and distributed brain regions has not been directly tested in humans. We therefore used targeted noninvasive electromagnetic stimulation to modulate human cortical-hippocampal networks and tested effects of this manipulation on memory. Multiple-session stimulation increased functional connectivity among distributed cortical-hippocampal network regions and concomitantly improved associative memory performance. These alterations involved localized long-term plasticity because increases were highly selective to the targeted brain regions, and enhancements of connectivity and associative memory persisted for ~24 hours after stimulation. Targeted cortical-hippocampal networks can thus be enhanced noninvasively, demonstrating their role in associative memory.

The ability to hold information in working memory is fundamental for cognition. Contrary to the long-standing view that working memory depends on sustained, elevated activity, we present evidence suggesting that humans can hold information in working memory via \"activity-silent\" synaptic mechanisms. Using multivariate pattern analyses to decode brain activity patterns, we found that the active representation of an item in working memory drops to baseline when attention shifts away. A targeted pulse of transcranial magnetic stimulation produced a brief reemergence of the item in concurrently measured brain activity. This reactivation effect occurred and influenced memory performance only when the item was potentially relevant later in the trial, which suggests that the representation is dynamic and modifiable via cognitive control. The results support a synaptic theory of working memory.

Memorization is an essential functionality that enables today's machine learning algorithms to provide a high quality of learning and reasoning for each prediction. Memorization gives algorithms prior knowledge to keep the context and define confidence for their decision. Unfortunately, the existing deep learning algorithms have a weak and non-transparent notion of memorization. Unlike existing deep learning models inspired by the brain's computing capability, HyperDimensional Computing (HDC) is introduced as a model of human memory. Therefore, it mimics several important functionalities of the brain memory by operating with a vector that is computationally tractable and mathematically rigorous in describing human cognition. In this manuscript, we introduce a brain-inspired system that represents HDC memorization capability over a graph of relations. We propose GrapHD, hyperdimensional memorization that represents graph-based information into high-dimensional space and enables reasoning. GrapHD defines an encoding method representing complex graph structure into high-dimensional space, supporting weighted and unweighted graphs. Our encoder spreads the information of all nodes and edges across into a full holistic representation so that no component is more responsible for storing any piece of information than another. Then, GrapHD defines several important cognitive functionalities over the encoded memory graph. These operations include memory reconstruction, information retrieval, graph matching, and shortest path. Our extensive evaluation shows that GrapHD: (1) significantly enhances learning capability by giving the notion of short/long term memorization to learning algorithms, (2) enables cognitive computing and reasoning over memorization graph, and (3) enables holographic brain-like computation with substantial robustness to noise and failure.

The neural mechanisms underlying memory regulation during sleep are not yet fully understood.We found that melanin concentrating hormone–producing neurons (MCH neurons) in the hypothalamus actively contribute to forgetting in rapid eye movement (REM) sleep. Hypothalamic MCH neurons densely innervated the dorsal hippocampus. Activation or inhibition of MCH neurons impaired or improved hippocampus-dependent memory, respectively. Activation of MCH nerve terminals in vitro reduced firing of hippocampal pyramidal neurons by increasing inhibitory inputs.Wake- and REM sleep–active MCH neurons were distinct populations that were randomly distributed in the hypothalamus. REM sleep state–dependent inhibition of MCH neurons impaired hippocampus-dependent memory without affecting sleep architecture or quality. REM sleep–active MCH neurons in the hypothalamus are thus involved in active forgetting in the hippocampus.

Alzheimer's disease (AD) manifests with memory loss and spatial disorientation. AD pathology starts in the entorhinal cortex, making it likely that local neural correlates of spatial navigation, particularly grid cells, are impaired. Grid-cell–like representations in humans can be measured using functional magnetic resonance imaging. We found that young adults at genetic risk for AD (APOE-ε4 carriers) exhibit reduced grid-cell–like representations and altered navigational behavior in a virtual arena. Both changes were associated with impaired spatial memory performance. Reduced grid-cell–like representations were also related to increased hippocampal activity, potentially reflecting compensatory mechanisms that prevent overt spatial memory impairment in APOE-ε4 carriers. Our results provide evidence of behaviorally relevant entorhinal dysfunction in humans at genetic risk for AD, decades before potential disease onset.

The focus of the present paper is on the contribution of the research in the student approaches to learning tradition. Several studies in this field have started from the assumption that students' approaches to learning develop towards more deep approaches to learning in higher education. This paper reports on a systematic review of longitudinal research on how students' approaches to learning develop during higher education. A total of 43 studies were included in the review. The results give an unclear picture of the development of approaches to learning and, thus, do not provide clear empirical evidence for the assumption that students develop towards more deep approaches during higher education. Neither methodological nor conceptual aspects of the studies investigated explained the ambiguity of the research results. Both theoretical and empirical implications for further research are discussed.