Automatic melody generation has recently gained significant attention in music creation and artificial intelligence. However, existing models often lack accuracy in emotional expression, coherence, and diversity. To address these issues, we propose a melody generation model based on Recurrent Neural Networks (RNN) and Variational Autoencoders (VAE), integrating emotional consistency loss and generative adversarial loss. This approach enhances melody diversity via VAE and captures long- and short-term dependencies using RNNs for better structural coherence. Emotional consistency loss helps maintain target emotions during generation, while generative adversarial loss improves naturalness and fluency. Experimental results show that our model outperforms traditional models like Music Transformer, MuseNet, and DeepBach in fluency, creativity, emotional expression, and harmony. The generated melodies are more expressive and innovative, providing a new method and perspective in melody generation, improving emotional expression and diversity, and laying a foundation for advancing automatic music creation technology.

Automatic melody generation has recently gained significant attention in music creation and artificial intelligence. However, existing models often lack accuracy in emotional expression, coherence, and diversity. To address these issues, we propose a melody generation model based on Recurrent Neural Networks (RNN) and Variational Autoencoders (VAE), integrating emotional consistency loss and generative adversarial loss. This approach enhances melody diversity via VAE and captures long- and short-term dependencies using RNNs for better structural coherence. Emotional consistency loss helps maintain target emotions during generation, while generative adversarial loss improves naturalness and fluency. Experimental results show that our model outperforms traditional models like Music Transformer, MuseNet, and DeepBach in fluency, creativity, emotional expression, and harmony. The generated melodies are more expressive and innovative, providing a new method and perspective in melody generation, improving emotional expression and diversity, and laying a foundation for advancing automatic music creation technology.

We propose a phase-assisted camera-in-the-loop optimization technique for holographic displays to enhance the quality of phase representation. Phase-only hologram is one of the most popular methods for three-dimensional hologram reproduction. However, the optical noise resulting from imperfections in the optical configuration, including the light source, the relay optics, and the spatial light modulator, affects the reconstructed hologram quality. We focus on constant noise terms that arise during the phase representation process in a phase-only spatial light modulator (PSLM). To estimate these noise terms, we employ a polarization-based 4-f system to acquire the constant complex term at the PSLM surface and the Fourier spectrum. Moreover, we use a random phase at iteration to effectively isolate and extract common noise components. Based on the extracted noise, we optimize the initial hologram using the modified angular spectrum propagator, taking into account the actual noise presented in the PSLM. Experimental results validate the feasibility of the proposed method, demonstrating that a single noise estimation significantly reduces speckle and improves image quality in multi-depth holograms.

We optimize computer-generated holograms for a spatial light modulator system that uses dual in-plane switching liquid crystal displays. Experiments demonstrate improved quality in the reconstructed image despite the limited modulation range.

This paper proposes a Quantum Fuzzy Support Vector Machine (QFSVM) model by integrating fuzzy set theory, quantum computing, and Support Vector Machines (SVM). A fuzzy membership function based on the K-nearest neighbor algorithm is introduced to reduce noise impact and differentiate sample importance. Quantum kernel methods replace traditional kernels to leverage quantum advantages. For stock market forecasting, the model incorporates stock factors and processes political and social news using FinBERT to extract news-related features. The QFSVM constructs a unique feature space, capturing complex nonlinear patterns in the data. Experimental results show that QFSVM outperforms SVM in accuracy, F1 score, and precision, achieving 76.55% accuracy—a 7.72% improvement over SVM—and surpassing state-of-the-art models. Ablation studies confirm the necessity of both quantum kernels and fuzzy membership functions. This research represents an innovative fusion of quantum computing, fuzzy set theory, and SVM in finance, advancing quantum machine learning and stock market prediction while offering new opportunities and challenges for future exploration.

Resolution limitation is a fundamental problem of digital holography. We propose a dual-sensor configuration and the feed-forward network-based super-resolution algorithm for geometric phase self-interference incoherent digital holography.

We proposed the resolution enhancement of self-interference incoherent digital holography using sparse acquisition. Stochastic gradient descent-based optimization is exploited to find and refine the absent holographic signals.