Cascade ranking architecture, composed of matching, pre-ranking, ranking and re-ranking stages, is usually adopted to balance the efficiency and effectiveness in real-world recommendation system (RS). As the middle stage of RS, pre-ranking aims to quickly filter out the low-quality items selected at the matching stage and then forwarding high-quality items to the ranking stage. Existing pre-ranking approaches mainly endure two problems 1) Sample Selection Bias (SSB) problem, which heavily limits the performance improvement of filtering out low-quality items owing to ignoring the data flow between stages; and 2) Ranking Consistency (RC) problem, which may cause the ranked lists of the ranking stage and previous pre-ranking stage to be inconsistent. As a result, the competitive items with high scores at the ranking stage may not be selected because of low scores at the pre-ranking stage. These both two problems may cause sub-optimal performances, but previous works usually only focus on the one of them. In this paper, we propose a novel Sample Debias and Ranking Consistency Joint Learning Framework (SDCL) to jointly alleviate SSB and RC problems. SDCL consists of two main modules including 1) Multi-Task Distillation Module (MTD), which enhances the ability of identifying high-quality items by distilling knowledge across all tasks simultaneously from the more complex ranking model which jointly trained with the pre-ranking model; and 2) Adaptive Negative Sample Learning Module (ANSL), which improves the performance of filtering out low-quality items by adaptively adjusting negative samples learning weights based on the current performance of model. SDCL seamlessly integrates two modules in an end-to-end multi-task learning framework. Evaluations on both real-world large-scale traffic logs and online A/B test demonstrate the efficacy and superiority of SDCL.

Human detection is an important problem which has many applications in collaborative manufacturing, such as employee counting and human fatigue detection. Recently, some methods based on convolutional neural network (CNN) have been proposed to solve this problem, but they often suffer from poor generalization due to the lack of real-world datasets from industrial scenarios. In this paper, we propose a graph pooling based human detection method to improve the generalization of existing CNN models. First, we segment an image and abstract it as a graph where nodes represent the segmented regions and the features of nodes are pixel values of the segmentation. Second, a graph pooling model is used to extract coarse-grained but informative features from the graph. Next, the pooled graph is reconstructed to image and input to a popular CNN model (i.e., YOLOv5) for human detection. We perform experiments on data containing a great number of images from industrial scenarios. Experimental results show that the proposed method outperforms the popular CNN model and has better generalization performance.