To solve the problem of gradient vanishing caused by too many layers of Convolutional Neural Network (CNN) in deep reinforcement learning, a deep reinforcement learning method based on weighted densely connected convolutional network was proposed. Firstly, image features were extracted by skip-connection structure in densely connected convolutional network. Secondly, weight coefficients were added into densely connected convolutional neural network, and each layer in a weighted densely connected convolutional network received all the feature maps generated by its previous layers and was initialized the weight in the skip-connection with different value. Finally, the weight of each layer was dynamically adjusted during training to extract features more effectively. Compared with conventional deep reinforcement learning, in GridWorld simulation experiment, the average reward value of the proposed method was increased by 85.67% under the same number of training steps; in FlappyBird simulation experiment, the average reward value was increased by 55.05%. The experimental results show that the proposed method can achieve better performance in game simulation experiments with different difficulty levels.

In order to solve the problem of not using the product heterogeneity well in recommendation algorithm, a modified mass diffusion algorithm was presented by considering the effect of the object popularity information on the user preference prediction. By introducing a tunable parameter of product popularity and simulating the mass diffusion process on the user-product bipartite network, the effect of the product popularity was quantitatively characterized. The experimental results on three empirical data sets which named MovieLens, Netflix and Last.FM show that, compared with the traditional mass diffusion method, the proposed algorithm can enhance the average ranking score by 25.6%, 10.96% and 1.2% respectively, and increase the diversity of the recommendation lists by 59.30%, 53.07% and 8.59% respectively. The proposed non-equilibrium mass diffusion algorithm can get more practical results.

Formal approaches are construction methods with accurate mathematical semantics, which are based on strict mathematical proofs. Generally, Petri nets are considered as a class of computation models to model the concurrent behavior. Also, formal specifications and analysis of a system can be conveniently developed by Petri nets. However, it is difficult to model a train control system with prototype Petri nets. The difficulties can be solved by extended Petri nets with inhibitor arcs. Hence, some key problems of train control systems were modeled and analyzed by the computation models of extended Petri nets in this paper. Two control sub-systems, station management sub-system and interval operation sub-system. were proposed. The former performed the entering and leaving of trains from stations by cooperative control. The later executed the safety control of block regions in stations, the safety recovery of emergency situations such as lightning stroke and the loss of signals, and the management of railway crossings. Finally, the activity, reachability, and boundedness of the proposed models were analyzed by S-invariants.