Article: Design of a cross-domain resource integration learning path generation model for innovative talent cultivation using bi-directional GAN and deep contrastive clustering network Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.58 pp.85 - 100 Abstract: To enhance the utilisation efficiency of interdisciplinary learning resources in cultivating innovative talents, this study proposes a fusion generative model integrating a bi-directional generative adversarial network (Bi-GAN) with a deep contrastive clustering network (DCCN). The model integrates multi-domain curriculum resource features via attention mechanism, uses Bi-GAN for feature analysis and enhancement, and finally applies DCCN to cluster and serialise resources into a coherent learning path. Experimental results show that: The silhouette coefficient, normalised mutual information, adjusted rand index, and path coherence score of the proposed model on the test set reach 0.36, 0.72, 0.56, and 0.78, respectively. Compared with the best results in the baseline methods, the proposed model achieves relative improvements of 24.1%, 10.8%, 14.3%, and 16.4% in SC, NMI, ARI, and PCS, respectively. Overall, the proposed model effectively realises deep cross-domain knowledge integration and coherent learning path generation, and provides a solution for personalised educational resource organisation. Inderscience Publishers - linking academia, business and industry through research

Title: Design of a cross-domain resource integration learning path generation model for innovative talent cultivation using bi-directional GAN and deep contrastive clustering network

Authors: Lian Tong; Liyan Zhou

Addresses: School of Information Technology, Jiangsu Maritime Institute, Nanjing 211170, China ' School of Information Technology, Jiangsu Maritime Institute, Nanjing 211170, China

Abstract: To enhance the utilisation efficiency of interdisciplinary learning resources in cultivating innovative talents, this study proposes a fusion generative model integrating a bi-directional generative adversarial network (Bi-GAN) with a deep contrastive clustering network (DCCN). The model integrates multi-domain curriculum resource features via attention mechanism, uses Bi-GAN for feature analysis and enhancement, and finally applies DCCN to cluster and serialise resources into a coherent learning path. Experimental results show that: The silhouette coefficient, normalised mutual information, adjusted rand index, and path coherence score of the proposed model on the test set reach 0.36, 0.72, 0.56, and 0.78, respectively. Compared with the best results in the baseline methods, the proposed model achieves relative improvements of 24.1%, 10.8%, 14.3%, and 16.4% in SC, NMI, ARI, and PCS, respectively. Overall, the proposed model effectively realises deep cross-domain knowledge integration and coherent learning path generation, and provides a solution for personalised educational resource organisation.

Keywords: two-way generation countermeasure network; deep contrast clustering; cross-domain resource integration; learning path generation; cultivation of innovative talents.

DOI: 10.1504/IJICT.2026.153805

International Journal of Information and Communication Technology, 2026 Vol.27 No.58, pp.85 - 100

Received: 20 Jan 2026
Accepted: 27 Feb 2026
Published online: 26 May 2026 *

Title: Design of a cross-domain resource integration learning path generation model for innovative talent cultivation using bi-directional GAN and deep contrastive clustering network

Keep up-to-date