No abstract

As a venue for the 2022 Winter Olympic and Paralympic Games, the National Alpine Skiing Center is located on Xiaohaitou Mountain in Beijing’s Yanqing district, where strong winds occur frequently. To reduce the impact of strong winds on the competition, windbreak fences have been installed in the ski area. To determine the effect and stability of the windbreak fences beside the ski slope of the alpine skiing center, the numerical simulation method was used to study the performance of 3.7 m, 4.7 m and 7.9 m high wind barriers. According to the actual meteorological conditions, two kinds of inlet wind speeds of 10 m/s and 33 m/s were set. The results show that the ambient wind speed is maximum only in a small area near the opening through the windbreak fence. When the ambient wind speed is 10 and 33 m/s, the wind speed in most drainage areas behind the barrier is below 5 and 15 m/s, respectively, which is significantly lower than the wind speed of incoming flow, and the wind protection effect is obvious. When the wind speed at the entrance height is 10 m and 33 m/s, the wind-proof effect is obvious. The wind-proof effect of 7.9 m high windproof bars is better than that of 3.7 and 4.7 m high windproof bars. The wind pressure at the top of the fence is the highest, and the wind pressure also increases when the wind speed increases. Under the action of maximum wind speed, the stability of a 7.9 m high storm fence is low.

Deep neural networks (DNNs) have achieved great success in the field of computer vision. The high requirements for memory and storage by DNNs make it difficult to apply them to mobile or embedded devices. Therefore, compression and structure optimization of deep neural networks have become a hot research topic. To eliminate redundant structures in deep convolutional neural networks (DCNNs), we propose an efficient filter pruning framework via deep reinforcement learning (DRL). The proposed framework is based on a deep deterministic policy gradient (DDPG) algorithm for filter pruning rate optimization. The main features of the proposed framework are as follows: (1) AA tailored reward function considering both accuracy and complexity of DCNN is proposed for the training of DDPG and (2) a novel filter sorting criterion based on Taylor expansion is developed for filter pruning selection. To illustrate the effectiveness of the proposed framework, extensive comparative studies on large public datasets and well-recognized DCNNs are conducted. The experimental results demonstrate that the Taylor-expansion-based filter sorting criterion is much better than the widely used minimum-weight-based criterion. More importantly, the proposed filter pruning framework can achieve over 10× parameter compression and 3× floating point operations (FLOPs) reduction while maintaining similar accuracy to the original network. The performance of the proposed framework is promising compared with state-of-the-art DRL-based filter pruning methods.