In recommender systems, the cold start problem, especially the user cold start one, is quite challenging yet important as it mostly comes with the rapid growth of the product and business. A well-performed new or early user recommender system has the potential to not only offer great onboarding experience, but also largely convert these users into loyal ones with considerable long-term values, which is strategically important from a product perspective. From a machine learning perspective, it is not an easy task as modern recommender systems heavily rely on large amounts of user-item interaction data to train a high quality model, and the data sparsity problem that new and early users bring is a clear challenge in terms of model training quality. In this short survey, among many proposed ideas on this topic, we focused on two general approaches for user cold start recommender system problems: metadata and meta-learning. Metadata methods generally aim at calculating similarities and learning representations by well utilizing limited user data that is collected regardless of users’ ratings for recommended items. Meta-learning methods learn the model in a more general and efficient way that could be adapted to new tasks with decent performance given data sparsity, which aligns the user cold start problem settings well.

Causation is gradually paid more attention to in industry as compared with correlation statement, it straightly targets on answering what-if questions, which generally delivers deeper and more insightful conclusions. Therefore, causal inference is naturally called. Mainly targeting on modeling counter-factual relationship that is usually not directly observable, causal inference has various of challenges on both problem setup and modeling side, which makes it a more complex topic than regular supervised learning task. As one of the heated discussed specific causal inference problems, conditional average treatment effect (CATE), or heterogeneous treatment effect (HTE), estimation model serves as a powerful tool in many applications, like personalized medicine and a series of uplift problems from user segmentation to ads budget optimization. Recently, several new CATE methods were proposed and we would like to do a short survey from the perspective of forest-based model to cover both meta-learners that could take random forest as base learner and forest-based specific CATE models. In total, we discussed 7 meta-learners and 5 forest-based specific models. We empirically evaluate these models with both synthetic data and real dataset.