RESUMO
We address a learning-to-normalize problem by proposing Switchable Normalization (SN), which learns to select different normalizers for different normalization layers of a deep neural network. SN employs three distinct scopes to compute statistics (means and variances) including a channel, a layer, and a minibatch. SN switches between them by learning their importance weights in an end-to-end manner. It has several good properties. First, it adapts to various network architectures and tasks (see Fig. 1). Second, it is robust to a wide range of batch sizes, maintaining high performance even when small minibatch is presented (e.g., 2 images/GPU). Third, SN does not have sensitive hyper-parameter, unlike group normalization that searches the number of groups as a hyper-parameter. Without bells and whistles, SN outperforms its counterparts on various challenging benchmarks, such as ImageNet, COCO, CityScapes, ADE20K, MegaFace and Kinetics. Analyses of SN are also presented to answer the following three questions: (a) Is it useful to allow each normalization layer to select its own normalizer? (b) What impacts the choices of normalizers? (c) Do different tasks and datasets prefer different normalizers? We hope SN will help ease the usage and understand the normalization techniques in deep learning. The code of SN has been released at https://github.com/switchablenorms.
RESUMO
The environmental-economic focus of wastewater treatment and management attracts growing attentions in recent years. The static efficiencies and their dynamic changes are helpful to systematically assess the environmental performance of the water agencies and wastewater treatment plants (WWTPs). Additionally, identifying key factors of efficiencies is critical to improve the operation of WWTPs. In this study, the window method of data envelopment analysis (DEA) was applied to estimate the annual efficiency for four Canadian WWTPs and to explore the variations of annual efficiency under different window lengths. Meanwhile, the Tobit regression analysis was developed to determine the driving forces for WWTPs' efficiency. The empirical results showed that: (i) the selected DEA window length remarkably affected both the average efficiency and the variations; however, it had no impact on the ranking of plants' efficiency; (ii) lower efficiencies were observed in plants with larger capacities due to higher infrastructure and operation investments involved; (iii) both the influent total phosphorus concentrations and influent flow rates had significant effects on the WWTPs' performance. Moreover, the staff and utility expenditures should be reduced to generate greater potential cost savings and efficiency improvement given the treatment technologies employed.