Modeling user's historical feedback is essential for Click-Through Rate Prediction in personalized search and recommendation. Existing methods usually only model users' positive feedback information such as click sequences which neglects the context information of the feedback. In this paper, we propose a new perspective for context-aware users' behavior modeling by including the whole page-wisely exposed products and the corresponding feedback as contextualized page-wise feedback sequence. The intra-page context information and inter-page interest evolution can be captured to learn more specific user preference. We design a novel neural ranking model RACP(Recurrent Attention over Contextualized Page sequence), which utilizes page-context aware attention to model the intra-page context. A recurrent attention process is used to model the cross-page interest convergence evolution as denoising the interest in the previous pages. Experiments on public and real-world industrial datasets verify our model's effectiveness.

Click-through rate (CTR) prediction plays an important role in search and recommendation, which are the two most prominent scenarios in e-commerce. A number of models have been proposed to predict CTR by mining user behaviors, especially users' interactions with items. But the sparseness of user behaviors is an obstacle to the improvement of CTR prediction. Previous works only focused on one scenario, either search or recommendation. However, on a practical e-commerce platform, search and recommendation share the same set of users and items, which means joint learning of both scenarios may alleviate the sparseness of user behaviors. In this paper, we propose a novel Search and Recommendation Joint Graph (SRJGraph) neural network to jointly learn a better CTR model for both scenarios. A key question of joint learning is how to effectively share information across search and recommendation, in spite of their differences. A notable difference between search and recommendation is that there are explicit queries in search, whereas no query exists in recommendation. We address this difference by constructing a unified graph to share representations of users and items across search and recommendation, as well as represent user-item interactions uniformly. In this graph, users and items are heterogeneous nodes, and search queries are incorporated into the user-item interaction edges as attributes. For recommendation where no query exists, a special attribute is attached on user-item interaction edges. We further propose an intention and upstream-aware aggregator to explore useful information from high-order connections among users and items. We conduct extensive experiments on a large-scale dataset collected from Taobao.com, the largest e-commerce platform in China. Empirical results show that SRJGraph significantly outperforms the state-of-the-art approaches of CTR prediction in both search and recommendation tasks.

Conversion rate (CVR) prediction is one of the most critical tasks for digital display advertising. Commercial systems often require to update models in an online learning manner to catch up with the evolving data distribution. However, conversions usually do not happen immediately after user clicks. This may result in inaccurate labeling, which is called delayed feedback problem. In previous studies, delayed feedback problem is handled either by waiting positive label for a long period of time, or by consuming the negative sample on its arrival and then insert a positive duplicate when conversion happens later. Indeed, there is a trade-off between waiting for more accurate labels and utilizing fresh data, which is not considered in existing works. To strike a balance in this trade-off, we propose Elapsed-Time Sampling Delayed Feedback Model (ES-DFM), which models the relationship between the observed conversion distribution and the true conversion distribution. Then we optimize the expectation of true conversion distribution via importance sampling under the elapsed-time sampling distribution. We further estimate the importance weight for each instance, which is used as the weight of loss function in CVR prediction. To demonstrate the effectiveness of ES-DFM, we conduct extensive experiments on a public data and a private industrial dataset. Experimental results confirm that our method consistently outperforms the previous state-of-the-art results.

Multi-scenario Learning to Rank is essential for Recommender Systems, Search Engines and Online Advertising in e-commerce portals where the ranking models are usually applied in many scenarios. However, existing works mainly focus on learning the ranking model for a single scenario, and pay less attention to learning ranking models for multiple scenarios. We identify two practical challenges in industrial multi-scenario ranking systems: (1) The Feedback Loop problem that the model is always trained on the items chosen by the ranker itself. (2) Insufficient training data for small and new scenarios. To address the above issues, we present ZEUS, a novel framework that learns a Zoo of ranking modEls for mUltiple Scenarios based on pre-training on users' spontaneous behaviors (e.g. queries which are directly searched in the search box and not recommended by the ranking system). ZEUS decomposes the training process into two stages: self-supervised learning based pre-training and fine-tuning. Firstly, ZEUS performs self-supervised learning on users' spontaneous behaviors and generates a pre-trained model. Secondly, ZEUS fine-tunes the pre-trained model on users' implicit feedback in multiple scenarios. Extensive experiments on Alibaba's production dataset demonstrate the effectiveness of ZEUS, which significantly outperforms state-of-the-art methods. ZEUS averagely achieves 6.0%, 9.7%, 11.7% improvement in CTR, CVR and GMV respectively than state-of-the-art method.

For better user satisfaction and business effectiveness, Click-Through Rate (CTR) prediction is one of the most important tasks in E-commerce. It is often the case that users' interests different from their past routines may emerge or impressions such as promotional items may burst in a very short period. In essence, such changes relate to item evolution problem, which has not been investigated by previous studies. The state-of-the-art methods in the sequential recommendation, which use simple user behaviors, are incapable of modeling these changes sufficiently. It is because, in the user behaviors, outdated interests may exist and the popularity of an item over time is not well represented. To address these limitations, we introduce time-aware item behaviors for addressing the recommendation of emerging preference. The time-aware item behavior for an item is a set of users who interact with this item with timestamps. The rich interaction information of users for an item may help to model its evolution. In this work, we propose a CTR prediction model TIEN based on the time-aware item behavior. In TIEN, by leveraging the interaction time intervals, information of similar users in a short time interval helps identify the emerging user interest of the target user. By using the sequential time intervals, the item's popularity over time can be captured in evolutionary item dynamics. Noisy users who interact with items accidentally are further eliminated thus learning robust personalized item dynamics. To the best of our knowledge, this is the first study to the item evolution problem for E-commerce CTR prediction. We conduct extensive experiments on five real-world CTR prediction datasets. The results show that the TIEN model consistently achieves remarkable improvements to the state-of-the-art methods.

Neuron-level structured pruning is a very effective technique to reduce the computation of neural networks without compromising prediction accuracy. In previous works, structured pruning is usually achieved by imposing L1 regularization on the scaling factors of neurons, and pruning the neurons whose scaling factors are below a certain threshold. The reasoning is that neurons with smaller scaling factors have weaker influence on network output. A scaling factor close to 0 actually suppresses a neuron. However, L1 regularization lacks discrimination between neurons because it pushes all scaling factors towards 0. A more reasonable pruning method is to only suppress unimportant neurons (with 0 scaling factors) and simultaneously keep important neurons intact (with larger scaling factor). To achieve this goal, we propose a new regularizer on scaling factors, namely polarization regularizer. Theoretically, we prove that polarization regularizer pushes some scaling factors to 0 and others to a value $a > 0$. Experimentally, we show that structured pruning using polarization regularizer achieves much better results than using L1 regularizer. Experiments on CIFAR and ImageNet datasets show that polarization pruning achieves the state-of-the-art result to date.

For better user experience and business effectiveness, Click-Through Rate (CTR) prediction has been one of the most important tasks in E-commerce. Although extensive CTR prediction models have been proposed, learning good representation of items from multimodal features is still less investigated, considering an item in E-commerce usually contains multiple heterogeneous modalities. Previous works either concatenate the multiple modality features, that is equivalent to giving a fixed importance weight to each modality; or learn dynamic weights of different modalities for different items through technique like attention mechanism. However, a problem is that there usually exists common redundant information across multiple modalities. The dynamic weights of different modalities computed by using the redundant information may not correctly reflect the different importance of each modality. To address this, we explore the complementarity and redundancy of modalities by considering modality-specific and modality-invariant features differently. We propose a novel Multimodal Adversarial Representation Network (MARN) for the CTR prediction task. A multimodal attention network first calculates the weights of multiple modalities for each item according to its modality-specific features. Then a multimodal adversarial network learns modality-invariant representations where a double-discriminators strategy is introduced. Finally, we achieve the multimodal item representations by combining both modality-specific and modality-invariant representations. We conduct extensive experiments on both public and industrial datasets, and the proposed method consistently achieves remarkable improvements to the state-of-the-art methods. Moreover, the approach has been deployed in an operational E-commerce system and online A/B testing further demonstrates the effectiveness.