On July 18, 2021, the PKU-DAIR Lab1)(data and Intelligence Research Lab at Peking University) openly released the source code of Hetu, a highly efficient and easy-to-use distributed deep learning(DL) framework. Hetu i...
On July 18, 2021, the PKU-DAIR Lab1)(data and Intelligence Research Lab at Peking University) openly released the source code of Hetu, a highly efficient and easy-to-use distributed deep learning(DL) framework. Hetu is the first distributed DL system developed by academic groups in Chinese universities, and takes into account both high availability in industry and innovation in academia. Through independent research and development, Hetu is completely decoupled from the existing DL systems and has unique characteristics. The public release of the Hetu system will help researchers and practitioners to carry out frontier MLSys(machine learning system) research and promote innovation and industrial upgrading.
Battery aging is a major hindrance to the advancement of batteries, particularly in their use for deep peaking and auxiliary services in the electricity market. Although existing literature recognizes battery aging in...
详细信息
Insect fine-grained image classification is an application scenario in fine-grained image classification. It not only has the characteristics of small inter-class differences and large intra-class differences, but als...
详细信息
The past decade has seen rapid growth of distributed stream data processing systems. Under these systems, a stream application is realized as a Directed Acyclic Graph (DAG) of operators, where the level of parallelism...
详细信息
The video grounding(VG) task aims to locate the queried action or event in an untrimmed video based on rich linguistic descriptions. Existing proposal-free methods are trapped in the complex interaction between video ...
详细信息
The video grounding(VG) task aims to locate the queried action or event in an untrimmed video based on rich linguistic descriptions. Existing proposal-free methods are trapped in the complex interaction between video and query, overemphasizing cross-modal feature fusion and feature correlation for VG. In this paper, we propose a novel boundary regression paradigm that performs regression token learning in a transformer. Particularly, we present a simple but effective proposal-free framework, namely video grounding transformer(ViGT), which predicts the temporal boundary using a learnable regression token rather than multi-modal or cross-modal features. In ViGT, the benefits of a learnable token are manifested as follows.(1) The token is unrelated to the video or the query and avoids data bias toward the original video and query.(2) The token simultaneously performs global context aggregation from video and query ***, we employed a sharing feature encoder to project both video and query into a joint feature space before performing cross-modal co-attention(i.e., video-to-query attention and query-to-video attention) to highlight discriminative features in each modality. Furthermore, we concatenated a learnable regression token [REG] with the video and query features as the input of a vision-language transformer. Finally, we utilized the token [REG] to predict the target moment and visual features to constrain the foreground and background probabilities at each timestamp. The proposed ViGT performed well on three public datasets:ANet-Captions, TACoS, and YouCookⅡ. Extensive ablation studies and qualitative analysis further validated the interpretability of ViGT.
To address the inefficiency of traditional methods for person-job matching and the lack of interpretability of deep learning approaches, a novel approach for person-job matching based on BM25 and pre-Trained language ...
详细信息
Colored Petri Nets (CPNs) provide descriptions of the concurrent behaviors for software and hardware. Model checking based on CPNs is an effective method to simulate and verify the concurrent behavior in system design...
详细信息
As one of the most important forensic tasks, reconstruction of the original information in tampered images is a key step for tampering detection and localization. Currently, a number of methods have been designed to e...
详细信息
High-temperature pre-stretching experiments were carried out on the AZ31 Mg alloy at 723 K with strain levels of 2.54%,6.48%,10.92%,and 19.2%to alter the microstructure and texture for improving room-temperature *** r...
详细信息
High-temperature pre-stretching experiments were carried out on the AZ31 Mg alloy at 723 K with strain levels of 2.54%,6.48%,10.92%,and 19.2%to alter the microstructure and texture for improving room-temperature *** results showed that the strain-hardening coefcient increased,while the Lankford value *** addition,the Erichsen values of all pre-stretch sheets were enhanced compared with that of the as-received *** maximum Erichsen value increased from 2.38 mm for the as-received sample to 4.03 mm for the 10.92%-stretched sample,corresponding to an improvement of 69.32%.This improvement was mainly attributed to the gradual increase in grain size,and the(0001)basal texture was weakened due to the activated non-basal slip as the high-temperature pre-stretching strain levels *** visco-plastic self-consistent analysis was performed on the as-received and high-temperature pre-stretched *** confrmed the higher activity of the prismatic slip in 10.92%-stretched sample,leading to divergence and weakening of basal texture *** results in an augmentation of the Schmid factor under diferent slip ***,it can be concluded that high-temperature pre-stretching technology provided an efective method to enhance the formability of Mg alloy sheets.
Long-term storage(LTS)can provide various services to address seasonal fluctuations in variable renewable energy by reducing energy ***,long-term unit commitment(UC)with LTS involves mixed-integer programming with lar...
详细信息
Long-term storage(LTS)can provide various services to address seasonal fluctuations in variable renewable energy by reducing energy ***,long-term unit commitment(UC)with LTS involves mixed-integer programming with large-scale coupling constraints between consecutive intervals(state-of-charge(SOC)constraint of LTS,ramping rate,and minimum up/down time constraints of thermal units),resulting in a significant computational ***,an iterative-based fast solution method is proposed to solve the long-term UC with ***,the UC with coupling constraints is split into several sub problems that can be solved in ***,the solutions of the sub problems are adjusted to obtain a feasible solution that satisfies the coupling ***,a decoupling method for long-term time-series coupling constraints is proposed to determine the global optimization of the SOC of the *** price-arbitrage model of the LTS determines the SOC boundary of the LTS for each sub ***,the sub problem with the SOC boundary of the LTS is iteratively solved *** proposed method was verified using a modified IEEE 24-bus *** results showed that the computation time of the unit combination problem can be reduced by 97.8%,with a relative error of 3.62%.
暂无评论