Scholarlink

<p>Qubit Neural Networks (QNNs) combine principles of quantum computing with artificial neural networks, offering a transformative boost to the efficiency and performance of deep learning models. In this study, we propose a hybrid model termed Quantum Small World Deep Echo State Network (Qubit SW Deep-ESN), which integrates qubit neurons into the reservoir of a Deep Echo State Network (Deep-ESN) designed with a Small World (SW) topology. The input data is first transformed into quantum states, enabling the reservoir’s neurons to leverage quantum mechanical properties such as superposition and entanglement. The SW topology enhances information propagation due to its short path lengths and clustered structure, further improving the model’s dynamic properties. Bayesian Optimization (BO) is employed to tune critical hyperparameters, ensuring optimal performance and stability. To evaluate the effectiveness of the proposed Qubit SW Deep-ESN, we conduct experiments on four Speech Emotion Recognition (SER) tasks: RAVDESS, Emo-DB, SAVEE, and TESS databases. The results demonstrate significant improvements in recognition accuracies (weighted and unweighted), achieving 79.83% and 78.36% on RAVDESS, 98% and 97% on Emo-DB, 75.45% and 75.07% on SAVEE, and 99.29% and 99.11% on TESS. These findings highlight the superior performance of the Qubit SW Deep-ESN compared to existing state-of-the-art approaches, showcasing its potential to advance neural information processing and improve emotion recognition capabilities.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantum-enhanced cortical deep echo state network for fast and accurate speech emotion recognition

  • 期刊论文

  • 发表日期：

    2025年8月2日

查看全文
  • Rebh Soltani,
  • Emna Benmohamed,
  • Hela Ltifi

摘要

Qubit Neural Networks (QNNs) combine principles of quantum computing with artificial neural networks, offering a transformative boost to the efficiency and performance of deep learning models. In this study, we propose a hybrid model termed Quantum Small World Deep Echo State Network (Qubit SW Deep-ESN), which integrates qubit neurons into the reservoir of a Deep Echo State Network (Deep-ESN) designed with a Small World (SW) topology. The input data is first transformed into quantum states, enabling the reservoir’s neurons to leverage quantum mechanical properties such as superposition and entanglement. The SW topology enhances information propagation due to its short path lengths and clustered structure, further improving the model’s dynamic properties. Bayesian Optimization (BO) is employed to tune critical hyperparameters, ensuring optimal performance and stability. To evaluate the effectiveness of the proposed Qubit SW Deep-ESN, we conduct experiments on four Speech Emotion Recognition (SER) tasks: RAVDESS, Emo-DB, SAVEE, and TESS databases. The results demonstrate significant improvements in recognition accuracies (weighted and unweighted), achieving 79.83% and 78.36% on RAVDESS, 98% and 97% on Emo-DB, 75.45% and 75.07% on SAVEE, and 99.29% and 99.11% on TESS. These findings highlight the superior performance of the Qubit SW Deep-ESN compared to existing state-of-the-art approaches, showcasing its potential to advance neural information processing and improve emotion recognition capabilities.