Under distribution shift (DS) where the training data distribution differs from the test one, a powerful technique is importance weighting (IW) which handles DS in two separate steps: weight estimation (WE) estimates the test-over-training density ratio and weighted classification (WC) trains the classifier from weighted training data. However, IW cannot work well on complex data, since WE is incompatible with deep learning. In this paper, we rethink IW and theoretically show it suffers from a circular dependency: we need not only WE for WC, but also WC for WE where a trained deep classifier is used as the feature extractor (FE). To cut off the dependency, we try to pretrain FE from unweighted training data, which leads to biased FE. To overcome the bias, we propose an end-to-end solution dynamic IW that iterates between WE and WC and combines them in a seamless manner, and hence our WE can also enjoy deep networks and stochastic optimizers indirectly. Experiments with two representative DSs on Fashion-MNIST and CIFAR-10/100 demonstrate that dynamic IW compares favorably with state-of-the-art methods.
A key assumption in supervised learning is that training and test data follow the same probability distribution. However, this fundamental assumption is not always satisfied in practice, e.g., due to changing environments, sample selection bias, privacy concerns, or high labeling costs. Transfer learning (TL) relaxes this assumption and allows us to learn under distribution shift. Classical TL methods typically rely on importance-weighting-a predictor is trained based on the training losses weighted according to the importance (i.e., the testover-training density ratio). However, as real-world machine learning tasks are becoming increasingly complex, high-dimensional, and dynamical, novel approaches are explored to cope with such challenges recently. In this article, after introducing the foundation of TL based on importance-weighting, we review recent advances based on joint and dynamic importancepredictor estimation. Furthermore, we introduce a method of causal mechanism transfer that incorporates causal structure in TL. Finally, we discuss future perspectives of TL research.
The macro policy of the stock market is an important market information. The implementation goal of the macro policy of the stock market is mainly to improve the effectiveness of the stock market. However, whether this effectiveness has achieved the goal is worth verifying through empirical data. The exertion of this information utility is closely related to the effectiveness of the stock market. Use the run test method in statistics to collect and sort out the daily data of stock price index in recent 30 years, the linkage between 75 macro policy events and 35 trading days of market efficiencies before and after the macro event are tested since 1992 to 2022. The results show that 50.66% of the macro policies are positively linked to the effectiveness of the stock market, while 49.34% of the macro policies have reduced the effectiveness of the market operation. This shows that the effectiveness of China’s stock market is not high, and the nonlinear characteristics are obvious, so the policy formulation of the stock market needs further improvement.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.