Code completion for programming education based on deep learning

Terada, Kenta; Watanobe, Yutaka

doi:10.1504/ijcistudies.2021.115424

Cited by 13 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After Zhang and Cao proposed the perceptron model, research on artificial neural networks began to increase, and there was a strong interest in this learning model of biological neural system models [ 21 ]. Terada and Watanobe trained deep neural network models with high-dimensional features after the introduction of deep neural networks with a complete representation of correlation information between features of sample data, combined with continuous feature information to form higher-order features and dimensional feature samples [ 22 ]. Doleck et al proposed a novel feedforward network learning algorithm for solving large-scale linear equations based on the introduction of the conjugate gradient method [ 23 ].…”

Section: Related Workmentioning

confidence: 99%

[Retracted] A Comprehensive Assessment of Cultivation Environment of Top Innovative High‐Level Talents Based on Deep Learning Algorithm

Zhu

Qin

2022

Journal of Environmental and Public Health

View full text Add to dashboard Cite

The quality of talent has increased across all fields due to the constant growth of different industries and the growing job saturation. Real-time job information on recruitment platforms can, therefore, accurately reflect the demand for talent from businesses, serving as a basis for the creation of training policies in schools. In international competition, the development of talents, especially top-level talents, will become more and more crucial. Growing in importance is China’s economy and social development. The evaluation of higher vocational and technical talents, however, should also be assessed from a variety of angles, given the diversification of talent training objectives and teaching methods, as well as the expansion of teaching functions. An emerging machine learning technology called deep learning (DL) has been developed to bring machine learning closer to the goals of artificial intelligence. This essay offers a thorough evaluation of the depth of deep learning as it relates to the development of innovative talent in schools. The entire school must be strengthened. It is demonstrated that the average execution time is slashed by 0.0024 s, and the learning sample size error of the DL model is reduced by 0.05276 when compared to the Apriori method. As a result, implementing and researching the DL model can significantly improve both the overall teaching quality of schools and their capacity for innovation.

show abstract

Section: Related Workmentioning

confidence: 99%

[Retracted] A Comprehensive Assessment of Cultivation Environment of Top Innovative High‐Level Talents Based on Deep Learning Algorithm

Zhu

Qin

2022

Journal of Environmental and Public Health

View full text Add to dashboard Cite

show abstract

“…In [28] a new framework based on the LSTM algorithm was proposed for performing personalized and adaptive learning, giving better results than the Deep Knowledge Tracing (DKT) method. And in [29], a method based on LSTM helps programming students by providing the next word following an incomplete program.…”

Section: Time-series Based Modelsmentioning

confidence: 99%

Using LSTM to Identify Help Needs in Primary School Scratch Students

Imbernón Cuadrado,

Manjarrés Riesco,

de la Paz López

2023

Applied Sciences

View full text Add to dashboard Cite

In the last few years, there has been increasing interest in the use of block-based programming languages as well as in the ethical aspects of Artificial Intelligence (AI) in primary school education. In this article, we present our research on the automatic identification of the need for assistance among primary school children performing Scratch exercises. For data collection, user experiences have been designed to take into account ethical aspects, including gender bias. Finally, a first-in-class distance calculation method for block-based programming languages has been used in a Long Short-Term Memory (LSTM) model, with the aim of identifying when a primary school student needs help while he/she carries out Scratch exercises. This model has been trained twice: the first time taking into account the gender of the students, and the second time excluding it. The accuracy of the model that includes gender is 99.2%, while that of the model that excludes gender is 91.1%. We conclude that taking into account gender in training this model can lead to overfitting, due to the under-representation of girls among the students participating in the experiences, making the model less able to identify when a student needs help. We also conclude that avoiding gender bias is a major challenge in research on educational systems for learning computational thinking skills, and that it necessarily involves effective and motivating gender-sensitive instructional design.

show abstract

“…Notably, the model exhibits high accuracy in discerning errors within faulty solution codes. Terada et al [16] presented an intriguing model designed for predicting subsequent code sequences to facilitate code completion. Leveraging an LSTM network architecture, their model serves as a valuable resource for novice programmers grappling with the challenge of crafting complete code from scratch.…”

Section: IImentioning

confidence: 99%

Transcompiler CNN-GMN Based Cross Language Clone Detection and Algorithm Recognition

Singh,

Bhutiya,

Goyal

et al. 2023

IJRASET

View full text Add to dashboard Cite

Software replication stemming from code reutilization introduces complexities in software maintenance. While deep learning-driven clone detection tools abound, they predominantly cater to singular programming languages. Concurrently, the task of identifying algorithms within program code remains challenging due to the absence of metadata, rendering the selection process intricate. This research seeks to establish synergies between these disparate domains by advancing innovative methodologies. The study delves into the unexplored domain of harnessing data augmentation, particularly leveraging transcompiler-based techniques, to enhance the discernment of code clones across different programming languages. Drawing upon the insights derived from transcompilers, this investigation applies data augmentation through source-to-source translation, exemplified by the Transcoder. This method extends the applicability of single-language models, such as the Graph Matching Network (GMN), to encompass cross-language detection. Concurrently, a program code classification model is introduced, leveraging Convolutional Neural Networks (CNNs) to discern algorithms based on structural features (SFs). These SFs are extracted from program codes and subsequently transformed into a one-hot binary matrix. The CNN model undergoes meticulous fine-tuning, optimizing its structural configurations and hyperparameters for superior algorithm classification. This research signifies an integration of disparate realms, intertwining cross-language clone detection and algorithm identification within a unified framework. The exploration of data augmentation for cross-language clone detection and the employment of CNN-GMNs for algorithm identification converge to furnish significant contributions to both domains. These insights, thus, hold promise for advancing the fields of software engineering and programming education

show abstract

Code completion for programming education based on deep learning

Cited by 13 publications

References 18 publications

[Retracted] A Comprehensive Assessment of Cultivation Environment of Top Innovative High‐Level Talents Based on Deep Learning Algorithm

[Retracted] A Comprehensive Assessment of Cultivation Environment of Top Innovative High‐Level Talents Based on Deep Learning Algorithm

Using LSTM to Identify Help Needs in Primary School Scratch Students

Transcompiler CNN-GMN Based Cross Language Clone Detection and Algorithm Recognition

Contact Info

Product

Resources

About