Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Lv, Zheqi, Ye, Keming, Wei, Zishu, Tian, Qi, Zhang, Shengyu, Zhang, Wenqiao, Wang, Wenjie, Kuang, Kun, Chua, Tat-Seng, Wu, Fei
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2501.07596
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912256553385984
author Lv, Zheqi
Ye, Keming
Wei, Zishu
Tian, Qi
Zhang, Shengyu
Zhang, Wenqiao
Wang, Wenjie
Kuang, Kun
Chua, Tat-Seng
Wu, Fei
author_facet Lv, Zheqi
Ye, Keming
Wei, Zishu
Tian, Qi
Zhang, Shengyu
Zhang, Wenqiao
Wang, Wenjie
Kuang, Kun
Chua, Tat-Seng
Wu, Fei
contents Deep neural networks have become foundational to advancements in multiple domains, including recommendation systems, natural language processing, and so on. Despite their successes, these models often contain incompatible parameters that can be underutilized or detrimental to model performance, particularly when faced with specific, varying data distributions. Existing research excels in removing such parameters or merging the outputs of multiple different pretrained models. However, the former focuses on efficiency rather than performance, while the latter requires several times more computing and storage resources to support inference. In this paper, we set the goal to explicitly improve these incompatible parameters by leveraging the complementary strengths of different models, thereby directly enhancing the models without any additional parameters. Specifically, we propose Compatibility-aware Knowledge Integration (CKI), which consists of Parameter Compatibility Assessment and Parameter Splicing, which are used to evaluate the knowledge content of multiple models and integrate the knowledge into one model, respectively. The integrated model can be used directly for inference or for further fine-tuning. We conduct extensive experiments on various datasets for recommendation and language tasks, and the results show that Compatibility-aware Knowledge Integration can effectively optimize incompatible parameters under multiple tasks and settings to break through the training limit of the original model without increasing the inference cost.
format Preprint
id arxiv_https___arxiv_org_abs_2501_07596
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimize Incompatible Parameters through Compatibility-aware Knowledge Integration
Lv, Zheqi
Ye, Keming
Wei, Zishu
Tian, Qi
Zhang, Shengyu
Zhang, Wenqiao
Wang, Wenjie
Kuang, Kun
Chua, Tat-Seng
Wu, Fei
Machine Learning
Computation and Language
Information Retrieval
Deep neural networks have become foundational to advancements in multiple domains, including recommendation systems, natural language processing, and so on. Despite their successes, these models often contain incompatible parameters that can be underutilized or detrimental to model performance, particularly when faced with specific, varying data distributions. Existing research excels in removing such parameters or merging the outputs of multiple different pretrained models. However, the former focuses on efficiency rather than performance, while the latter requires several times more computing and storage resources to support inference. In this paper, we set the goal to explicitly improve these incompatible parameters by leveraging the complementary strengths of different models, thereby directly enhancing the models without any additional parameters. Specifically, we propose Compatibility-aware Knowledge Integration (CKI), which consists of Parameter Compatibility Assessment and Parameter Splicing, which are used to evaluate the knowledge content of multiple models and integrate the knowledge into one model, respectively. The integrated model can be used directly for inference or for further fine-tuning. We conduct extensive experiments on various datasets for recommendation and language tasks, and the results show that Compatibility-aware Knowledge Integration can effectively optimize incompatible parameters under multiple tasks and settings to break through the training limit of the original model without increasing the inference cost.
title Optimize Incompatible Parameters through Compatibility-aware Knowledge Integration
topic Machine Learning
Computation and Language
Information Retrieval
url https://arxiv.org/abs/2501.07596