Memory capacity is a primary determinant of learning rate
the verdict
CONTESTED
contested - evenly split
refutedsupported
the weight of evidence
2 sources for · 1 against
The retrieved sources present mixed evidence, with some computational studies indicating capacity matters for memorization while research on working memory points to selective attention as a primary determinant of learning ability.
This paper studies how the model architecture and data configurations influence the empirical memorization capacity of generative transformers. The models are trained using synthetic text datasets derived from the Systematized Nomenclature of Medicine (SNOMED) knowledge graph: triplets, representing static connections, and sequences, simulating complex relation patterns. The results show that embedding size is the primary determinant of learning speed and capacity, while additional layers provide limited benefits and may hinder performance on simpler datasets. Activation functions play a crucial role, and Softmax demonstrates greater stability and capacity. Furthermore, increasing the complexity of the data set seems to improve the final memorization. These insights improve our understanding of transformer memory mechanisms and provide a framework for optimizing model design with structured real-world data.
A single factor (i.e., general intelligence) can account for much of an individuals’ performance across a wide variety of cognitive tests. However, despite this factor’s robustness, the underlying process is still a matter of debate. To address this question, we developed a novel battery of learning tasks to assess the general learning abilities (GLAs) of mice. Using this battery, we previously reported a strong relationship between GLA and a task designed to tax working memory capacity (i.e., resistance to competing demands). Here we further explored this relationship by investigating which aspects of working memory (storage or processing) best predict GLAs in mice. We found that a component of working memory, selective attention, correlated with GLA comparably to working memory capacity. However, this relationship was not found for two other components of working memory, short-term memory capacity and duration. These results provide further evidence that variations in aspects of working memory and executive functions covary with general cognitive abilities.
Developing strong mathematical skills at the start of primary school is vital for children’s later learning and development. Educational maths apps delivered on touch-screen tablet devices are suited to primary education and an emerging evidence base demonstrates their potential to support young children’s early mathematical development. The educational maths apps at the focus of thesis draw on the principles of active, engaged, meaningful, and socially interactive learning combined with curriculum-based content and specific learning goals. Using a mixed-methods, pragmatic approach and a multi-level, ecological, determinant framework, this thesis aimed to address how does app-based mathematics instruction work, who does it work for, and under what circumstances does it work? The UK Proof of Concept study (Chapter 4) showed children aged 4-5 years identified as low-achievers in mathematics (n = 12) who used the apps for 8 weeks made greater learning gains in mathematics compared to their typically attaining peers receiving standard mathematical practice (n = 15). No significant effect of children’s socio-economic status was found but children with a poor memory capacity demonstrated stronger learning gains with the apps. Similar results were found in the Brazil Proof of Concept study (Chapter 5). After a 10-week intervention period, results showed children aged 5-6 years made greater progress in mathematics with the apps when delivered in their first (Brazilian Portuguese, n =
Everything we examined (3)
This check searched the claim as stated. It did not run a separate search for evidence against it.