QoSDiff: An Implicit Topological Embedding Learning Framework Leveraging Denoising Diffusion and Adversarial Attention for Robust QoS Prediction

TL;DR

QoSDiff addresses scalable QoS prediction without explicit user--service graphs by combining a single-step diffusion-based embedding learning module (DELM) with an adversarial bidirectional attention interaction module (AAIM). The diffusion component denoises context-enhanced embeddings to produce robust latent representations, while the adversarial attention framework captures high-order user--service dependencies and suppresses noisy observations. Empirical results on WS-DREAM and the large-scale EEL dataset show that QoSDiff outperforms strong baselines, with improved cross-dataset generalization and robustness to observational noise. The work demonstrates that generative diffusion in latent spaces, coupled with adversarial interaction modeling, offers a scalable and resilient alternative to graph neural networks for QoS prediction in dynamic service ecosystems.

Abstract

Accurate Quality of Service (QoS) prediction is fundamental to service computing, providing essential data-driven guidance for service selection and ensuring superior user experiences. However, prevalent approaches, particularly Graph Neural Networks (GNNs), heavily rely on constructing explicit user--service interaction graphs. Such reliance not only leads to the intractability of explicit graph construction in large-scale scenarios but also limits the modeling of implicit topological relationships and exacerbates susceptibility to environmental noise and outliers. To address these challenges, this paper introduces \emph{QoSDiff}, a novel embedding learning framework that bypasses the prerequisite of explicit graph construction. Specifically, it leverages a denoising diffusion probabilistic model to recover intrinsic latent structures from noisy initializations. To further capture high-order interactions, we propose an adversarial interaction module that integrates a bidirectional hybrid attention mechanism. This adversarial paradigm dynamically distinguishes informative patterns from noise, enabling a dual-perspective modeling of intricate user--service associations. Extensive experiments on two large-scale real-world datasets demonstrate that QoSDiff significantly outperforms state-of-the-art baselines. Notably, the results highlight the framework's superior cross-dataset generalization capability and exceptional robustness against observational noise.

Figures (8)

• Figure 1: QoSDiff can effectively learn embeddings for QoS prediction even in the absence of explicit graph structures.
• Figure 2: Illustration of the QoS prediction task in QoSDiff. Observed user--service interactions with known QoS values $r_{ij}$ and context attributes are used to train the prediction model $f_\theta$. The trained model then estimates the missing QoS values $\hat{r}_{ij}$ for unseen user--service pairs $(i,j)\in\bar{\Omega}$.
• Figure 3: Overall architecture of the proposed QoSDiff framework.
• Figure 4: Model performance comparison of different embedding learning methods
• Figure 5: Model performance comparison of different interaction learning methods