[2603.26138] PruneFuse: Efficient Data Selection via Weight Pruning and Network Fusion

Computer Science > Machine Learning arXiv:2603.26138 (cs) [Submitted on 27 Mar 2026] Title:PruneFuse: Efficient Data Selection via Weight Pruning and Network Fusion Authors:Humaira Kousar, Hasnain Irshad Bhatti, Jaekyun Moon View a PDF of the paper titled PruneFuse: Efficient Data Selection via Weight Pruning and Network Fusion, by Humaira Kousar and 2 other authors View PDF HTML (experimental) Abstract:Efficient data selection is crucial for enhancing the training efficiency of deep neural networks and minimizing annotation requirements. Traditional methods often face high computational costs, limiting their scalability and practical use. We introduce PruneFuse, a novel strategy that leverages pruned networks for data selection and later fuses them with the original network to optimize training. PruneFuse operates in two stages: First, it applies structured pruning to create a smaller pruned network that, due to its structural coherence with the original network, is well-suited for the data selection task. This small network is then trained and selects the most informative samples from the dataset. Second, the trained pruned network is seamlessly fused with the original network. This integration leverages the insights gained during the training of the pruned network to facilitate the learning process of the fused network while leaving room for the network to discover more robust solutions. Extensive experimentation on various datasets demonstrates that PruneFuse significa...