This paper investigates how agents learn and perform efficient strategies by trying different actions in asymmetric confrontation setting. Firstly, we use Gomoku as an example to analyse the causes and impacts of asymmetric confrontation: the first mover gains higher power than the second mover. We find that the first mover learns how to attack quickly while it is difficult for the second mover to learn how to defend since it cannot win the first mover and always receives negative rewards. As such, the game is stuck at a deadlock in which the first mover cannot make further advances to learn how to defend, and the second mover learns nothing. Secondly, we propose an ancillary mechanism (AM) to add two principles to the agent’s actions to overcome this difficulty. AM is a guidance for the agents to reduce the learning difficulty and to improve their behavioural quality. To the best of our knowledge, this is the first study to define asymmetric confrontation in reinforcement learning and propose approaches to tackle such problems. In the numerical tests, we first conduct a simple human vs AI experiment to calibrate the learning process in asymmetric confrontation. Then, an experiment of 15*15 Gomoku game by letting two agents (with AM and without AM) compete is applied to check the potential of AM. Results show that adding AM can make both the first and the second movers become stronger in almost the same amount of calculation.

Edge detection is one of the most fundamental fields in computer vision. With the rapid development of the combination of Convolutional Neural Network and Multi-Scale Representation of image, significant progress has been made in this field. However, most of them have a huge size, which makes it hard to apply in reality, and a huge number of parameters may lead to waste of computing resources. In this paper, we focus on qualitative analysis of the role of each part in the network, and propose a modified light-weight architecture based on our result and the study of former works. Our new architecture is composed of residual-blocks, max-pooling layers and batch normalization layers. Compared with the previous models, the new architecture performs better in memory, convergence and computation efficiency with similar model size. Moreover, the new architecture can achieve better accuracy with smaller model size. When evaluating our model on the well-known BSDS500 benchmark, we achieve ODS F-measure of 0.769 with parameters less than 0.3[Formula: see text]M, which shows a better property than the state-of-the-art result 0.766 at this level.