Song Qing

Learn More
For training algorithms of recurrent neural networks (RNN), convergent speed and training error are always two contradictory performances. In this letter, we propose a normalized adaptive recurrent learning (NARL) to obtain a tradeoff between transient and steady-state response. An augmented term is added to error gradient to exactly model the derivative of(More)
The Rapidly-exploring Random Tree (RRT) is a useful path planning algorithm and has been extensively researched in recent years. Till now parameters setting of the RRT algorithm have not yet been explored and are usually set based on the expert experience. In this paper, lots of simulation experiments are conducted for different parameter values. The(More)
Many applications of natural language processing (NLP) need an accurate resolution of various ambiguities existing in natural language. The task of fulfilling this need is also called word sense disambiguation (WSD). WSD is to resolve the correct sense for an instance of a polysemous word. On the other hand, as one of the most popular machine learning(More)
For the problem of unmanned aerial vehicle route planning in unknown environment, a rolling rapidly-exploring random tree algorithm is proposed. According to the current environment information, the local route planning is carried out in the rolling window. Unmanned aerial vehicle flight in accordance with the planning results. At the same time, the new(More)
An improved bidirectional RRT algorithm is proposed to solve the problem of path re-planning in a dynamic environment. The off-line path planning is carried out before the UAV takes off. During flight, if pop-up threat is detected, nodes which are affected by pop-up threat will be deleted and the remaining tree structure will be retained, then improved(More)
Chaotic neural networks have been successfully applied in pattern association problems in many research. However there are few in-depth theoretical analysis for such networks, such as stability issues. In this paper, we propose a new type of chaotic recurrent neural network (CRNN) which is more powerful in pattern association comparing to previous work.(More)
  • 1