Deep learning has been widely applied to digital art and music creation. However, producing melodies that follow music theory and match human compositional patterns remains challenging. This study proposes a symbolic music generation system that integrates supervised learning and reinforcement learning. The core framework employs recurrent neural networks for sequence modeling, while a reinforcement learning module formulates music rules as reward functions to guide pitch, duration, and rhythm. This hybrid approach helps produce outputs that better match human compositional patterns. We deploy the proposed framework as an Internet-based system that supports five distinct music styles and is publicly accessible online.
Over the past five years since the enactment of the environmental protection tax, a measure has been implemented that has harmonized the conflicting interests among the government, E-businesses, and the E-market. A substantial body of game-theoretic analysis has focused exclusively on the interactions between government entities, producers (firms), and the general populace (consumers). This paper introduces an OSP (Online Supervision Platform) into the game framework, which assists government in sharing ecological environmental guardianship responsibilities, concurrently stimulating society and the E-market to explore environmental protection tax loopholes. It constructs a four-party evolutionary game model. The analysis assesses the stability of strategic choices for each participant in the game. Furthermore, a stability analysis of potential equilibrium points in the system is conducted based on the first law of Lyapunov stability. Simulation delves into the impact of key elements on the E-commerce's evolution, scientifically examines the influence of OSP on stakeholders during the environmental protection tax implementation, providing a more comprehensive decision-making basis for environmental protection.
Debris flows occur abruptly and possess strong destructive power, making them one of the most hazardous geomorphological processes capable of inflicting heavy losses in a short period of time. In view of the difficulty in obtaining timely and accurate predictions using existing warning techniques, this study introduces a novel time-series forecasting strategy for debris-flow hazards, built upon an enhanced LSTM model. Field surveys were first conducted to assemble the initial dataset, after which the data were carefully filtered and standardized. Principal Component Analysis was applied to reorganize and condense the numerous environmental and geological variables, allowing eight dominant factors related to debris-flow development to be extracted. Based on these refined inputs, an LSTM-driven prediction architecture was developed. The model was trained and validated using the prepared dataset, and its performance was benchmarked against several traditional forecasting methods. Experimental outcomes reveal that the proposed model maintains low prediction errors-below 0.06 for training, 0.12 for testing, and 0.19 for validation. These results highlight the method's superior predictive capability and robustness compared with conventional approaches. The framework presented here has the potential to significantly enhance debris-flow risk assessment and offers practical support for emergency management and early-warning operations.
The integrity of data greatly affects its quality, and thus its asset value and also the accuracy of data analysis tasks based on it. However, in reality, there are often various factors that cause data deficiency. Although many scholars have made researches on this problem, most of the studies have limited accuracy. Therefore, this paper proposes a transformer-GRU based GAN, which first learns the potential distribution of real data through generator formed by stacked encoder of transformer to generate the complementary data, and then inputs the complementary data into the discriminator based on GRU for discrimination. Finally, it trains the model through the adversarial game between generator and discriminator. Comparison experiments are carried out based on the sales data of a company, and it can be seen that the evaluation indicators of the method proposed in this paper are superior to those of the compared methods, proving that the missing data recovery method proposed in this paper has better accuracy.
Given the swift evolution of oceanic surveillance systems and the substantial expansion of diverse marine intelligent devices and services, the serious attenuation of the acoustic channels in underwater sensor networks and the scarce computing resources of the ocean surface have led to the inability to meet the requirements of Unmanned Surface vehicle (USV) users for low latency and low energy consumption. This paper constructs an unmanned aerial vehicle (UAV)-assisted multi-access computing offloading model in ocean monitoring networks. Specifically, in the data collection stage, Underwater sensor nodes (USNs) transmit the collected data to the USVs via non-orthogonal multiple access (NOMA). In the data processing stage, USVs offloads part of the data to the UAV via frequency division multiple access (FDMA), and the UAV and USVs cooperate to complete the data processing task. In order to meet the requirements of USVs, this paper introduces a joint optimization problem, which involves optimizing the USN's transmission power, USV's transmission power, USV's offloading decision, and USV's bandwidth allocation. Considering the independence between the two stages, the optimization problem is decomposed into two separate stage-problems. The optimization problem of the data collection stage enables collected data to arrive at the USV simultaneously by using the particle swarm optimization algorithm to control the USN's transmission power. The optimization problem of the data processing stage is decomposed into three sub-problems: the USV's transmission power, USV's offloading decision, and USV's bandwidth allocation, which are optimized separately by using gradient descent, genetic algorithm, and CVX solver respectively. Through iterative alternation, the optimal solution of the data processing stage can be obtained. The simulation results demonstrate a considerable decrease in sum of latency and energy consumption for USVs with the proposed algorithm.
