Fusing Memory and Attention: A study on LSTM, Transformer and Hybrid Architectures for Symbolic Music Generation
This work addresses the challenge of generating coherent symbolic music for applications in music composition and AI-assisted creativity, though it is incremental as it builds on existing LSTM and Transformer methods.
This paper tackled the problem of symbolic music generation by comparing LSTM and Transformer architectures, finding that LSTMs capture local patterns but fail at long-range dependencies, while Transformers model global structure but produce irregular phrasing; the proposed hybrid architecture combining a Transformer Encoder with an LSTM Decoder achieved better local and global continuity and coherence, as shown by evaluations on 1,000 generated melodies using 17 musical quality metrics.
Machine learning techniques, such as Transformers and Long Short-Term Memory (LSTM) networks, play a crucial role in Symbolic Music Generation (SMG). Existing literature indicates a difference between LSTMs and Transformers regarding their ability to model local melodic continuity versus maintaining global structural coherence. However, their specific properties within the context of SMG have not been systematically studied. This paper addresses this gap by providing a fine-grained comparative analysis of LSTMs versus Transformers for SMG, examining local and global properties in detail using 17 musical quality metrics on the Deutschl dataset. We find that LSTM networks excel at capturing local patterns but fail to preserve long-range dependencies, while Transformers model global structure effectively but tend to produce irregular phrasing. Based on this analysis and leveraging their respective strengths, we propose a Hybrid architecture combining a Transformer Encoder with an LSTM Decoder and evaluate it against both baselines. We evaluated 1,000 generated melodies from each of the three architectures on the Deutschl dataset. The results show that the hybrid method achieves better local and global continuity and coherence compared to the baselines. Our work highlights the key characteristics of these models and demonstrates how their properties can be leveraged to design superior models. We also supported the experiments with ablation studies and human perceptual evaluations, which statistically support the findings and provide robust validation for this work.