Multi-domain learning (MDL) has become a prominent topic in enhancing the quality of personalized services. It's critical to learn commonalities between domains and preserve the distinct characteristics of each domain. However, this leads to a challenging dilemma in MDL. On the one hand, a model needs to leverage domain-aware modules such as experts or embeddings to preserve each domain's distinctiveness. On the other hand, real-world datasets often exhibit long-tailed distributions across domains, where some domains may lack sufficient samples to effectively train their specific modules. Unfortunately, nearly all existing work falls short of resolving this dilemma. To this end, we propose a novel Cross-experts Covariance Loss for Disentangled Learning model (Crocodile), which employs multiple embedding tables to make the model domain-aware at the embeddings which consist most parameters in the model, and a covariance loss upon these embeddings to disentangle them, enabling the model to capture diverse user interests among domains. Empirical analysis demonstrates that our method successfully addresses both challenges and outperforms all state-of-the-art methods on public datasets. During online A/B testing in Tencent's advertising platform, Crocodile achieves 0.72% CTR lift and 0.73% GMV lift on a primary advertising scenario. The code is openly accessible at: https://github.com/SkylerLinn/Crocodile.

Click-through rate (CTR) prediction is a crucial area of research in online advertising. While binary cross entropy (BCE) has been widely used as the optimization objective for treating CTR prediction as a binary classification problem, recent advancements have shown that combining BCE loss with an auxiliary ranking loss can significantly improve performance. However, the full effectiveness of this combination loss is not yet fully understood. In this paper, we uncover a new challenge associated with the BCE loss in scenarios where positive feedback is sparse: the issue of gradient vanishing for negative samples. We introduce a novel perspective on the effectiveness of the auxiliary ranking loss in CTR prediction: it generates larger gradients on negative samples, thereby mitigating the optimization difficulties when using the BCE loss only and resulting in improved classification ability. To validate our perspective, we conduct theoretical analysis and extensive empirical evaluations on public datasets. Additionally, we successfully integrate the ranking loss into Tencent's online advertising system, achieving notable lifts of 0.70% and 1.26% in Gross Merchandise Value (GMV) for two main scenarios. The code is openly accessible at: https://github.com/SkylerLinn/Understanding-the-Ranking-Loss.

The exponential growth of Ethereum transactions has resulted in a significant increase in phishing scams, leading to substantial financial losses in recent years. Current machine/deep learning-based approaches for classification have been found to be inadequate for large-scale and label-imbalanced Ethereum scenarios. To address this issue, we propose Phish2vec, a novel network embedding approach that takes into account the transaction temporality and heterogeneity in detecting phishing scams on Ethereum. Our approach begins by producing a transaction sub-network through data collection and preprocessing, which includes a novel Statistics-Based Sampling (SBS) method to address label leakage. To generate sequences that contain more comprehensive information, we then utilize two different types of sequences generators: Temporal-based Sequences Generator (TSG) and Heterogeneous-based Sequences Generator (HSG). By concatenating the sequences generated by TSG and HSG together, and feeding them into Word2vec and Fully Connected neural network (FC), our approach can identify phishing accounts with an Fl-score as high as 82.05%, which significantly outperforms classic schemes such as DeepWalk (67.29%), Trans2vec (74.78%), and Node2vec (70.91%).