Federated Deep Knowledge Tracing

Wu, Jinze; Huang, Zhenya; Liu, Qi; Lian, Defu; Wang, Hao; Chen, Enhong; Ma, Haiping; Wang, Shijin

doi:10.1145/3437963.3441747

Cited by 14 publications

(10 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In addition, we include some recently proposed approaches to cover a very wide range of DLKT models with different design focus. Please note that there are a few newly developed DLKT focusing on either utilizing more auxiliary information such as LPKT that utilizes time spent on questions [29], or solving data isolation problems in KT via federated learning [42]. The generalization of these approaches are limited to specific datasets and are out of scope in this paper.…”

Section: Representative Dlkt Methodsmentioning

confidence: 99%

pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models

Liu¹,

Liu²,

Huang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Knowledge tracing (KT) is the task of using students' historical learning interaction data to model their knowledge mastery over time so as to make predictions on their future interaction performance. Recently, remarkable progress has been made of using various deep learning techniques to solve the KT problem. However, the success behind deep learning based knowledge tracing (DLKT) approaches is still left somewhat mysterious and proper measurement and analysis of these DLKT approaches remain a challenge. First, data preprocessing procedures in existing works are often private and/or custom, which limits experimental standardization. Furthermore, existing DLKT studies often differ in terms of the evaluation protocol and are far away real-world educational contexts. To address these problems, we introduce a comprehensive python based benchmark platform, PYKT, to guarantee valid comparisons across DLKT methods via thorough evaluations. The PYKT library consists of a standardized set of integrated data preprocessing procedures on 7 popular datasets across different domains, and 10 frequently compared DLKT model implementations for transparent experiments. Results from our fine-grained and rigorous empirical KT studies yield a set of observations and suggestions for effective DLKT, e.g., wrong evaluation setting may cause label leakage that generally leads to performance inflation; and the improvement of many DLKT approaches is minimal compared to the very first DLKT model proposed by Piech et al. [25]. We have open sourced PYKT and our experimental results at https://pykt.org/. We welcome contributions from other research groups and practitioners.

show abstract

Section: Representative Dlkt Methodsmentioning

confidence: 99%

pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models

Liu¹,

Liu²,

Huang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Moreover, the learning of high-quality KT models inevitably requires a substantial amount of data to guarantee training stability. However, practical educational scenarios often suffer from the cold-start problem and the data isolation problem: e.g., students' learning data tends to be distributed across different schools and is also highly proprietary, so that it is difficult to gather the data for training [116]. Therefore, potential methods of combining concepts such as federated learning or active learning to train novel KT models are also a promising research direction.…”

Section: Knowledge Tracing With Less Learning Datamentioning

confidence: 99%

A Survey of Knowledge Tracing

Liu¹,

Shen²,

Huang³

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

High-quality education is one of the keys to achieving a more sustainable world. The recent COVID-19 epidemic has triggered the outbreak of online education, which has enabled both students and teachers to learn and teach at home. Meanwhile, it is now possible to record and research a large amount of learning data using online learning platforms in order to offer better intelligent educational services. Knowledge Tracing (KT), which aims to monitor students' evolving knowledge state, is a fundamental and crucial task to support these intelligent services. Therefore, an increasing amount of research attention has been paid to this emerging area and considerable progress has been made. In this survey, we propose a new taxonomy of existing basic KT models from a technical perspective and provide a comprehensive overview of these models in a systematic manner. In addition, many variants of KT models have been proposed to capture more complete learning process. We then review these variants involved in three phases of the learning process: before, during, and after the student learning, respectively. Moreover, we present several typical applications of KT in different educational scenarios. Finally, we provide some potential directions for future research in this fast-growing field.

show abstract

“…However, deep knowledge tracing still faces challenges because the data island problem also exists in the field of education. In order to solve the challenges of data scarcity, data imbalance, and difficult data comparison caused by data isolation, J. Wu et al (2021) introduced the federated learning framework into deep knowledge tracing to train deep knowledge tracing models of each educational island to improve the quality of the model. And their study is the first work in this field.…”

Section: Applications Of Federated Learning In Data Miningmentioning

confidence: 99%

A survey on federated learning in data mining

Mao

et al. 2021

WIREs Data Min & Knowl

View full text Add to dashboard Cite

Data mining is a process to extract unknown, hidden, and potentially useful information from data. But the problem of data island makes it arduous for people to collect and analyze scattered data, and there is also a privacy security issue when mining data. A collaboratively decentralized approach called federated learning unites multiple participants to generate a shareable global optimal model and keeps privacy‐sensitive data on local devices, which may bring great hope to us for solving the problems of decentralized data and privacy protection. Though federated learning has been widely used, few systematic studies have been conducted on the subject of federated learning in data mining. Hence, different from prior reviews in this field, we make a comprehensive summary and provide a novel taxonomy of the application of federated learning in data mining. This article starts by providing a thorough description of the relevant definitions and concepts, followed by an in‐depth investigation on the challenges faced by federated learning. In this context, we elaborate four taxonomies of major applications of federated learning in data mining, including education, healthcare, IoT, and intelligent transportation, and discuss them comprehensively. Finally, we discuss four promising research directions for further research, that is, privacy enhancement, improvement of communication efficiency, heterogeneous system processing, and reducing economic costs. This article is categorized under: Technologies > Machine Learning Fundamental Concepts of Data and Knowledge > Big Data Mining Technologies > Artificial Intelligence

show abstract

Federated Deep Knowledge Tracing

Cited by 14 publications

References 27 publications

pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models

pyKT: A Python Library to Benchmark Deep Learning based Knowledge Tracing Models

A Survey of Knowledge Tracing

A survey on federated learning in data mining

Contact Info

Product

Resources

About