This work proposes a model reference adaptive control based on recursive neural networks. This secondary-level controller corrects the deviations on the voltage and frequency setpoints of a simple primary control in a...
详细信息
Flexible and transparent conductive films that can be processed at room temperature are strongly required. We have obtained the In2O3-based transparent thin-films with both conductivity and flexibility for next-genera...
详细信息
Amorphous oxide semiconductors have attracted attention because of their low-temperature processability and high field-effect mobility. However, hysteresis remains in most case when the TFT channel is deposited at roo...
详细信息
We fabricated SnO films from SnO2 target using a combined process of reductive sputtering and annealing at 800 °C. However, annealing for 30 min resulted in rough surface and precipitation of Sn due to disproport...
详细信息
Few‐shot image classification is the task of classifying novel classes using extremely limited labelled *** perform classification using the limited samples,one solution is to learn the feature alignment(FA)informati...
详细信息
Few‐shot image classification is the task of classifying novel classes using extremely limited labelled *** perform classification using the limited samples,one solution is to learn the feature alignment(FA)information between the labelled and unlabelled sample *** FA methods use the feature mean as the class prototype and calculate the correlation between prototype and unlabelled features to learn an alignment ***,mean prototypes tend to degenerate informative features because spatial features at the same position may not be equally important for the final classification,leading to inaccurate correlation ***,the authors propose an effective intraclass FA strategy that aggregates semantically similar spatial features from an adaptive reference prototype in low‐dimensional feature space to obtain an informative prototype feature map for precise correlation ***,a dual correlation module to learn the hard and soft correlations was developed by the *** module combines the correlation information between the prototype and unlabelled features in both the original and learnable feature spaces,aiming to produce a comprehensive cross‐correlation between the prototypes and unlabelled *** both FA and cross‐attention modules,our model can maintain informative class features and capture important shared features for *** results on three few‐shot classification benchmarks show that the proposed method outperformed related methods and resulted in a 3%performance boost in the 1‐shot setting by inserting the proposed module into the related methods.
To realize a carbon-neutral society, integrating distributed energy resources (DERs), such as photovoltaic systems (PVs) and electric vehicles (EVs), into distribution networks is imperative. As PV generations and EV ...
详细信息
SnSO4 has a layered structure and is a promising candidate as a p-type transparent semiconductor with a band gap of 3.9 eV and hole effective mass of 0.88, theoretically. However, there has been no report on experimen...
详细信息
Video embedding is the pivot in Temporal Action Detection (TAD). Once the video embedding can robustly capture the essence of actions and perceive activities in complex scenes, the TAD model can more accurately locali...
详细信息
This paper discusses energy and cost comparison for 9 different combinations of Photovoltaic (PV) and Lithium-Ion Battery Energy Storage System (BESS) sizes with load demand peak changes for the Kayangel Power System ...
详细信息
The utilization of renewable energy (RE) has increased worldwide because it is one of the effective ways to minimize greenhouse gases. The power outputs of photovoltaic (PV) and wind farms depend on weather conditions...
详细信息
暂无评论