Knowledge tracing is a crucial task that involves modeling learners’ knowledge levels and predicting their future learning performance. However, traditional deep knowledge tracing approaches often overlook the intrinsic relationships among learning features, treating them equally and failing to align with real learning scenarios. To address these issues, this paper proposes the multiple learning features, enhanced knowledge tracing (MLFKT) framework. Firstly, we construct learner–resource response (LRR) channels based on psychometric theory, establishing stronger intrinsic connections among learning features and overcoming the limitations of the item response theory. Secondly, we leverage stacked auto-encoders to extract low-dimensional embeddings for different LRR channels with denser representations. Thirdly, considering the varying impact of different LRR channels on learning performance, we introduce an attention mechanism to assign distinct weights to each channel. Finally, to address the challenges of memory retention and forgetting in the learning process and to handle long-term dependency issues, we employ a bidirectional long short-term memory network to model learners’ knowledge states, enabling accurate prediction of learning performance. Through extensive experiments on two real datasets, we demonstrate the effectiveness of our proposed MLFKT approach, which outperforms six traditional methods. The newly proposed method can enhance educational sustainability by improving the diagnosis of learners’ self-cognitive structures and by empowering teachers to intervene and personalize their teaching accordingly.

/pdf/multiple-learning-features-enhanced-knowledge-tracing-based-1526fnqu.pdf

Multiple Learning Features–Enhanced Knowledge Tracing Based on Learner–Resource Response Channels

Knowledge Tracking (KT) predicts student performance by modeling the mastery of knowledge components in past interactions. Although the latest research on KT is excellent to model behavior changes, the increasingly complex knowledge components and semantic relationships of new exercises have made courses sparse and difficult to adequately diagnose knowledge mastery. Hence, it is not enough to satisfy only the personalized learning. This paper proposes a graph-based method with multi-hierarchical network (Caps-HAGKT) for tracking capsule knowledge, which integrates rich structural information of the knowledge space, including multiple skills, prerequisites, attribution, and conceptual meta-paths. Firstly, we use the multi-hierarchical knowledge capsule network to track a variety of high-level knowledge concepts, and then extract the latent knowledge structure between levels. In addition, we adopt a new graph network with an external memory matrix to model the relationship between concept capsules at the same level and update concept states. In previous works, it was difficult to use complex graph information without any expert pre-definition. In Caps-HAGKT, the semantic attention of knowledge capsules is used to merge different semantics in knowledge according to their importance. Extensive experiments on some real-world datasets show that the prediction performance of Caps-HAGKT is better than that of the existing KT method (AUC is increased by 3.05% on an average). The visualization of both latent graphs and case studies indicates the excellent interpretability of predictions.

Modeling knowledge proficiency using multi-hierarchical capsule graph neural network

Plastic gating network: Adapting to personal development and individual differences in knowledge tracing

The development of intelligent tutoring system has greatly influenced the way students learn and practice, which increases their learning efficiency. The intelligent tutoring system must model learners' mastery of the knowledge before providing feedback and advices to learners, so one class of algorithm called "knowledge tracing" is surely important. This paper proposed Deep Self-Attentive Knowledge Tracing (DSAKT) based on the data of PTA, an online assessment system used by students in many universities in China, to help these students learn more efficiently. Experimentation on the data of PTA shows that DSAKT outperforms the other models for knowledge tracing an improvement of AUC by 2.1% on average, and this model also has a good performance on the ASSIST dataset.

Application of Deep Self-Attention in Knowledge Tracing.

Knowledge tracing (KT) is based on modeling students’ behavior sequences to obtain students’ knowledge state and predict students’ future performance. The KT task aims to model students’ knowledge state in real-time according to their historical learning behavior, so as to predict their future learning performance. Online education has become more critical in recent years due to the impact of COVID-19, and KT has also attracted much attention due to its importance in the education field. However, previous KT models generally have the following three problems. Firstly, students’ learning and forgetting behaviors affect their knowledge state, and past KT models have yet to exploit this fully. Secondly, the input of traditional KT models is mainly limited to students’ exercise sequence and answers. In the learning process, students’ answering performance can reflect their knowledge level. Finally, the context of students’ learning sequence also affects their judgment of the knowledge state. In this paper, we combined educational psychology theories to propose enhanced learning and forgetting behavior for contextual knowledge tracing (LFEKT). LFEKT enriches the features of exercises by introducing difficulty information and considers the influence of students’ answering behavior on the knowledge state. In order to model students’ learning and forgetting behavior, LFEKT integrates multiple influencing factors to build a knowledge acquisition module and a knowledge retention module. Furthermore, LFEKT introduces a long short-term memory (LSTM) network to capture the contextual relations of learned sequences. From the experimental results, it can be seen that LFEKT had better prediction performance than existing models on four public datasets, which indicates that LFEKT can better trace students’ knowledge state and has better prediction performance.

/pdf/enhanced-learning-and-forgetting-behavior-for-contextual-3pgalvk5.pdf

Enhanced Learning and Forgetting Behavior for Contextual Knowledge Tracing

Knowledge tracing is an important research topic in student modeling. The aim is to model a student's knowledge state by mining a large number of exercise records. The dynamic key-value memory network (DKVMN) proposed for processing knowledge tracing tasks is considered to be superior to other methods. However, through our research, we have noticed that the DKVMN model ignores both the students' behavior features collected by the intelligent tutoring system (ITS) and their learning abilities, which, together, can be used to help model a student's knowledge state. We believe that a student's learning ability always changes over time. Therefore, this article proposes a new exercise record representation method, which integrates the features of students' behavior with those of the learning ability, thereby improving the performance of knowledge tracing. Our experiments show that the proposed method can improve the prediction results of DKVMN.

/pdf/dynamic-key-value-memory-networks-with-rich-features-for-1zfhti83vm.pdf

Dynamic Key-Value Memory Networks With Rich Features for Knowledge Tracing.

Knowledge tracing is a significant research topic in student modeling. It is a task to model students' mastery level of knowledge points by mining their historical exercise performance. Literature has shown that Dynamic Key-Value Memory Networks (DKVMN) which has been proposed to handle the knowledge tracing task generally outperform traditional methods. However, through our experimentation, we have noticed a problem in the DKVMN model that it ignored behavior features collected by intelligence tutoring system (ITS) but only regard the exercise and the correctness as input. Behavior features, such as the response time, the hint request and the number of attempt, can be used to capture the student's learning behavior information and are very helpful in modeling the student's knowledge status. Therefore, the performance of the model can be improved. This work aims to improve the performance of the DKVMN model by incorporating more features to the input. More specifically, we apply decision tree classifier to preprocess the behavior features, which is an effective way to capture how the student deviates from others in the exercise. The predicted response concatenated with the exercise tag to train a DKVMN model, which can output the probability that a student will answer the exercise correctly. The experiment results show that our adapted DKVMN model, incorporating more combinations of behavior features can effectively improve accuracy. On the ASSISTments 2009 education dataset, the AUC value of our experiment is eight percent higher than the original DKVMN model.

Muti-behavior features based knowledge tracking using decision tree improved DKVMN

Although deep convolution neural networks (DC-NNs) can achieve remarkable success on medical image segmentation, their performance might significantly deteriorate when confronting testing data with the new distribution. Recent studies suggest that one major cause of this issue is the strong inductive bias of DCNNs, which towards image styles (e.g., superficial texture) that are sensitive to change, instead of the invariant content (e.g., object shapes). Inspired by this, we propose a novel method, named Invariant Content Synergistic Learning (ICSL), to improve the generalization ability of DCNNs on unseen data by controlling the inductive bias. Specifically, ICSL first mixes the style of training instances to perturb the training distribution, so that more diverse domains or styles would be made available for training DCNNs. Then, based on the perturbed distribution, we carefully design a dual-branches invariant content synergistic learning strategy to prevent style-biased predictions and maintain the invariant content. Extensive experimental results demonstrate the superior performance of the proposed method over state-of-the-art domain generalization methods on two typical medical segmentation tasks.

Invariant Content Synergistic Learning for Domain Generalization on Medical Image Segmentation

Deep convolutional neural networks (DCNNs) significantly improve the performance of medical image segmentation. Nevertheless, medical images frequently experience distribution discrepancies, which fails to maintain their robustness when applying trained models to unseen clinical data. To address this problem, domain generalization methods were proposed to enhance the generalization ability of DCNNs. Feature space-based data augmentation methods have proven their effectiveness to improve domain generalization. However, existing methods still mainly rely on certain prior knowledge or assumption, which has limitations in enriching the diversity of source domain data. In this paper, we propose a random feature augmentation (RFA) method to diversify source domain data at the feature level without prior knowledge. Specifically, we explore the effectiveness of random convolution at the feature level for the first time and prove experimentallyt hat itc an adequately preserve domain-invariant information while perturbing domainspecific information. Furthermore, tocapture the same domain-invariant information from the augmented features of RFA, we present a domain-invariant consistent learning strategy to enable DCNNs to learn a more generalized representation. Our proposed method achieves state-of-the-art performance on two medical image segmentation tasks, including optic cup/disc segmentation on fundus images and prostate segmentation on MRI images.

Xu Zhao

Papers

Dynamic Key-Value Memory Networks With Rich Features for Knowledge Tracing.

Muti-behavior features based knowledge tracking using decision tree improved DKVMN

Invariant Content Synergistic Learning for Domain Generalization on Medical Image Segmentation

A Random Feature Augmentation for Domain Generalization in Medical Image Segmentation