Research on Polyphonic Music Generation Algorithm Based on GPT Large Model

Edge Along with the rapid technological progress in the field of artificial intelligence, music generation algorithms based on large-scale pre-trained models have increasingly become the focus of academic attention. Existing polyphonic music generation techniques have limitations in terms of melodic complexity and harmonic diversity. To address this issue, this paper introduced an innovative polyphonic music generation scheme, the core of which is the Transformer architecture based on the GPT pre-training model. The algorithm refines the model parameters through a fine-tuning mechanism in order to efficiently capture long-term dependencies and complex harmonic patterns in musical sequences for high-quality polyphonic music composition. Specifically, targeted fine-tuning operations are implemented to enhance the performance of the GPT model for polyphonic music understanding and generation. The self-attention mechanism and positional coding technique were integrated to deepen the model’s understanding of the complex dependencies between musical sequences, and to enhance its accuracy and fluency in generating polyphonic music. The experimental results show that the method demonstrates significant advantages in objective evaluation dimensions, such as note accuracy and harmonic consistency, and also gains positive feedback in subjective evaluation dimensions, including listener satisfaction and the overall fluency of the music. With the help of cutting-edge deep learning technology and in-depth musicological evidence, the method proposed in this paper opens up innovative paths in the field of polyphonic music composition, and significantly demonstrates the great potential of this technology in practical scenarios.