CAN: Feature Co-Action for Click-Through Rate Prediction

Bian, Weijie; Wu, Kailun; Ren, Lejian; Pi, Qi; Zhang, Yujing; Chu, Xiao; Sheng, Xiang-Rong; Zhu, Yong-Nan; Chan, Zhangming; Mou, Na; Luo, Xinchen; Xiang, Shiming; Zhou, Guorui; Zhu, Xiaoqiang; Deng, Hongbo

doi:10.48550/arxiv.2011.05625

Cited by 4 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is worth noting that our generation process is in instance-level. Therefore, for the convenience of description, we rewrite Equation (5). Assuming our current batch size is 𝐵, then 𝒁 could be rewritten as:…”

Section: Parameters Generationmentioning

confidence: 99%

“…In the multi-domain discriminator, AFT utilizes knowledge representation learning to model relations between items, users, and domain indicators. (5). From the perspective of the attention mechanism.…”

Section: Multi-domain Recommendationmentioning

confidence: 99%

“…Dynamic weight generation approaches have also been widely used in the context of recommendation systems. CAN [5] proposes a co-action unit in which the items model's weights are generated dynamically. APG [40] generates different models' weight matrices and gets great performance.…”

Section: Dynamic Weight Generation In Deep Neural Networkmentioning

confidence: 99%

See 2 more Smart Citations

HAMUR: Hyper Adapter for Multi-Domain Recommendation

Li,

Yan,

Zhao

et al. 2023

Proceedings of the 32nd ACM International Conference on Information and Knowledge Management

View full text Add to dashboard Cite

Multi-Domain Recommendation (MDR) has gained significant attention in recent years, which leverages data from multiple domains to enhance their performance concurrently. However, current MDR models are confronted with two limitations. Firstly, the majority of these models adopt an approach that explicitly shares parameters between domains, leading to mutual interference among them. Secondly, due to the distribution differences among domains, the utilization of static parameters in existing methods limits their flexibility to adapt to diverse domains. To address these challenges, we propose a novel model Hyper Adapter for MUlti-Domain Recommendation (HAMUR). Specifically, HAMUR consists of two components: (1). Domain-specific adapter, designed as a pluggable module that can be seamlessly integrated into various existing multidomain backbone models, and (2). Domain-shared hyper-network, which implicitly captures shared information among domains and dynamically generates the parameters for the adapter. We conduct extensive experiments on two public datasets using various backbone networks. The experimental results validate the effectiveness and scalability of the proposed model. And we release our code implementation publicly 12 .

show abstract

Section: Parameters Generationmentioning

confidence: 99%

Section: Multi-domain Recommendationmentioning

confidence: 99%

See 1 more Smart Citation

HAMUR: Hyper Adapter for Multi-Domain Recommendation

Li,

Yan,

Zhao

et al. 2023

Proceedings of the 32nd ACM International Conference on Information and Knowledge Management

View full text Add to dashboard Cite

show abstract

“…Deep learning based methods have achieved great success in CTR prediction task [6,9,18,27]. Recently, a series of works [2,8,28] on modeling user behavior sequence have emerged. DIN [32], DIEN [31], MIND [15] and Transformer [25] usually model short-term user behaviors due to the limitation of latency.…”

Section: Related Workmentioning

confidence: 99%

Attentive capsule network for click-through rate and conversion rate prediction in online advertising

Chen

et al. 2021

Knowledge-Based Systems

View full text Add to dashboard Cite

“…The multi-feature combination allows linear models to learn nonlinear Feature 14 . By combining individual features with the Cartesian product, each feature will correspond to a hidden variable, thus deriving multiple input features to represent nonlinear relationships 15 .…”

Section: Introductionmentioning

confidence: 99%

Enhanced Efficiency and Security in Cross-Chain Transmission of Blockchain Internet of Ports through Multi-Feature-Based Joint Learning

Xie,

Zhang,

Liu

2024

Preprint

View full text Add to dashboard Cite

To enhance the timeliness of cross-chain transmission (CCT) in port blockchain Internet of Things (IoT), this study proposes a load adaptive control method tailored for CCT within port blockchain IoT networks. It establishes a balanced configuration model for CCT links within the blockchain IoT framework in port areas. The method facilitates compensation and adjustment of joint autocorrelation dynamic parameters pertinent to blockchain IoT devices within port environments through load-balancing scheduling techniques. Furthermore, it examines the correlation constraint parameter characteristics governing CCT control of blockchain IoT within port zones. Leveraging parameter estimation and priority scheduling methodologies for blockchain IoT CCT links, the study employs multi-feature joint learning (MFJL) to construct a load estimation model specific to CCT within port blockchain IoT settings. Moreover, the interference suppression of blockchain IoT CCT in port areas is addressed via the MFJL algorithm, which also facilitates the analysis of distinct characteristic components of cross-chain switching channels. The study implements adaptive multi-channel joint bus control and link balance designs for blockchain IoT CCT within port locales through grid scheduling and cross-chain parameter fusion methods. These endeavours aim to enhance the balance and real-time performance of blockchain IoT CCT within port zones. Simulation outcomes demonstrate notable efficacy, with the average input value of cross-chain energy efficiency for blockchain IoT in port areas reaching 5Mbps, the highest output value peaking at 9Mbps, and the average CCT queue length hovering around 1dB. Additionally, the energy consumption of blockchain IoT devices in port areas is observed to decrease by 1KJ by adopting this methodology.

show abstract

CAN: Feature Co-Action for Click-Through Rate Prediction

Cited by 4 publications

References 0 publications

HAMUR: Hyper Adapter for Multi-Domain Recommendation

HAMUR: Hyper Adapter for Multi-Domain Recommendation

Attentive capsule network for click-through rate and conversion rate prediction in online advertising

Enhanced Efficiency and Security in Cross-Chain Transmission of Blockchain Internet of Ports through Multi-Feature-Based Joint Learning

Contact Info

Product

Resources

About