Evolvable Psychology Informed Neural Network for Memory Behavior Modeling

TL;DR

Memory behavior modeling faces a trade-off between interpretability and predictive power. PsyINN addresses this by jointly optimizing a temporal neural network with differential sparse regression that evolves memory equations via descriptor evolution using differentiating operators. The architecture comprises a Deep Learning Module, a Differential Sparse Regression Module, and a Buffer Queue Module, trained with the objective $L_T = L_{Data} + \alpha L_{SR}$, where $L_{Data}$ and $L_{SR}$ are the data-loss and sparse-regression losses. Experiments on four large-scale real-world datasets show PsyINN surpassing state-of-the-art baselines and yielding interpretable, evolving symbolic memory equations through the SR module. Ablation and application studies demonstrate improved robustness and potential to guide memory research in psychology and education.

Abstract

Memory behavior modeling is a core issue in cognitive psychology and education. Classical psychological theories typically use memory equations to describe memory behavior, which exhibits insufficient accuracy and controversy, while data-driven memory modeling methods often require large amounts of training data and lack interpretability. Knowledge-informed neural network models have shown excellent performance in fields like physics, but there have been few attempts in the domain of behavior modeling. This paper proposed a psychology theory informed neural networks for memory behavior modeling named PsyINN, where it constructs a framework that combines neural network with differentiating sparse regression, achieving joint optimization. Specifically, to address the controversies and ambiguity of descriptors in memory equations, a descriptor evolution method based on differentiating operators is proposed to achieve precise characterization of descriptors and the evolution of memory theoretical equations. Additionally, a buffering mechanism for the sparse regression and a multi-module alternating iterative optimization method are proposed, effectively mitigating gradient instability and local optima issues. On four large-scale real-world memory behavior datasets, the proposed method surpasses the state-of-the-art methods in prediction accuracy. Ablation study demonstrates the effectiveness of the proposed refinements, and application experiments showcase its potential in inspiring psychological research.

Figures (7)

• Figure 1: (a) shows vocabulary flashcards in MaiMemo. (b) provides an example of learners alternating between different vocabulary words. (c) illustrates our research motivation: using traditional theoretical equations to inspire neural networks, while neural networks can optimize traditional theoretical equations, achieving a synergistic effect
• Figure 2: Input Feature Map
• Figure 3: Model Structure Diagram
• Figure 4: Training Error Trend of Alternating Gradient Descent Strategy on Duolingo and MaiMemo Datasets
• Figure 5: Change in Test Metrics with Different Selection Strategies on Duolingo and MaiMemo Datasets