a b s t r a c tAs a type of clean and renewable energy source, wind power is widely used. However, owing to the uncertainty of wind speed, it is essential to build an accurate forecasting model for large-scale wind power penetration. Numerical weather prediction (NWP) and data-driven modeling are two typical paradigms. NWP is usually unavailable or spatially insufficient. Data-driven modeling is an effective candidate. As to some newly-built wind farms, sufficient historical data is not available for training an accurate model, while some older wind farms may have long-term wind speed records. A question arises regarding whether the prediction model trained by data coming from older farms is also effective for a newly-built farm. In this paper, we propose an interesting trial of transferring the information obtained from data-rich farms to a newly-built farm. It is well known that deep learning can extract a high-level representation of raw data. We introduce deep neural networks, trained by data from data-rich farms, to extract wind speed patterns, and then finely tune the mapping with data coming from newly-built farms. In this way, the trained network transfers information from one farm to another. The experimental results show that prediction errors are significantly reduced using the proposed technique.
Real-world data often follows a long-tailed distribution, which makes the performance of existing classification algorithms degrade heavily. A key issue is that the samples in tail categories fail to depict their intra-class diversity. Humans can imagine a sample in new poses, scenes and view angles with their prior knowledge even if it is the first time to see this category. Inspired by this, we propose a novel reasoning-based implicit semantic data augmentation method to borrow transformation directions from other classes. Since the covariance matrix of each category represents the feature transformation directions, we can sample new directions from similar categories to generate definitely different instances. Specifically, the long-tailed distributed data is first adopted to train a backbone and a classifier. Then, a covariance matrix for each category is estimated, and a knowledge graph is constructed to store the relations of any two categories. Finally, tail samples are adaptively enhanced via propagating information from all the similar categories in the knowledge graph. Experimental results on CIFAR-LT-100, ImageNet-LT, and iNaturalist 2018 have demonstrated the effectiveness of our proposed method compared with the state-of-the-art methods.
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.