Traceability system is one of the popular applications of graphic blockchains. However, there are centralisation problems and a long time for final consistency confirmation in the graphic blockchain. In addition, the blockchain system in the traceability application scenario has the problem of insufficient supervision. Therefore, a witness‐based graphic blockchain consensus mechanism is proposed. In the consensus mechanism, a verifiable random function is used to screen the publishers of the unit; an SM2 threshold signature is used to sign the unit information to improve the non‐repudiation of the traceability information uploaders to the unit information under the supervision of the witness. The improved consistency algorithm cancels the process of finding a stable main chain and makes relatively many nodes to participate in the consensus process. The experimental results show that the graphic blockchain using the improved witness mechanism can reduce the degree of centralisation, shorten the time for new units to reach consensus, and greatly ensure the security and scalability of the blockchain system.

Online social networks greatly promote peoples' online interaction, where trust plays a crucial role. Trust prediction with trust path search is widely used to help users find the trusted friends and obtain valid information. However, the shortcomings of accuracy and time still exist in some famous algorithms. Therefore, the dynamic bidirectional heuristic search (DBHS) algorithm is proposed in this paper to find the reliable trust path by studying the heuristic search. First, the trust value and path length are comprehensively considered to find the most trusted user. Specially, it constrains the traversal depth based on the ‘small world’ theory and obtains the acceptable path set by using the relaxation coefficient λ to relax the depth of the shortest path. By this way, some longer path with the higher trust can be considered to improve the precision of algorithm. Then, an adjustment factor is designed based on the meet in the middle (MM) algorithm to assign search weights to two directions based on the size of the search tree expanded, so as to improve the problem of no priori when fixed parameters are used. Besides, the complexity of unidirectional trust path search can also be reduced by searching from two directions, which can reduce the depth and improve the efficiency of search. Finally, the predictive trust degree is outputted by the trust propagation function. Two public datasets are used to generate experimental results, which show that DBHS can quickly search and form reliable trust relationship, and it partly improves other algorithms.

Many methods based on deep learning have achieved amazing results in image sentiment analysis. However, these existing methods usually pursue high accuracy, ignoring the effect on model training efficiency. Considering that when faced with large‐scale sentiment analysis tasks, the high accuracy rate often requires long experimental time. In view of the weakness, a method that can greatly improve experimental efficiency with only small fluctuations in model accuracy is proposed, and singular value decomposition (SVD) is used to find the sparse feature of the image, which are sparse vectors with strong discriminativeness and effectively reduce redundant information; The authors propose the Fast Dictionary Learning algorithm (FDL), which can combine neural network with sparse representation. This method is based on K‐Singular Value Decomposition, and through iteration, it can effectively reduce the calculation time and greatly improve the training efficiency in the case of small fluctuation of accuracy. Moreover, the effectiveness of the proposed method is evaluated on the FER2013 dataset. By adding singular value decomposition, the accuracy of the test suite increased by 0.53%, and the total experiment time was shortened by 8.2%; Fast Dictionary Learning shortened the total experiment time by 36.3%.

This study proposes a new generation network based on transformers and guided by the music theory to produce high‐quality music work. In this study, the decoding block of the transformer is used to learn the internal information of single‐track music, and cross‐track transformers are used to learn the information amongst the tracks of different musical instruments. A reward network based on the music theory is proposed, which optimizes the global and local loss objective functions while training and discriminating the network so that the reward network can provide a reliable adjustment method for the generation of the network. The method of combining the reward network and cross entropy loss is used to guide the training of the generator and produce high‐quality music work. Compared with other multi‐track music generation models, the experimental results verify the validity of the model.