WiMi Hologram Cloud Inc. announced the development of a technology--Quantum Semi-Supervised Learning (QSSL) achieved through quantum supremacy. The core idea of this technology is to leverage the unique advantages of quantum computing, enabling effective learning even with limited labeled data. By harnessing the extraordinary speed of quantum computing, it overcomes the limitations of classical computing.
The protocol proposed by WiMi is not only applicable to quantum semi-supervised learning but can also be extended to other quantum machine learning tasks, thereby advancing the widespread practical application of quantum computing. Quantum Semi-Supervised Learning is the quantum version of classical semi-supervised learning. The key to classical semi-supervised learning lies in the ability to train models using a combination of a small amount of labeled data and a large amount of unlabeled data.
In this process, the parallelism and superposition properties of quantum computing allow the algorithm to process massive amounts of unlabeled data in a very short time, significantly improving both model training efficiency and prediction accuracy. Compared to traditional methods, the quantum self-training algorithm not only offers speed advantages but also can handle more complex and high-dimensional datasets. Another key technology developed by WiMi is the Quantum Semi-Supervised K-Means Clustering Algorithm.
In traditional K-means clustering, data points are dividend into K cluster centers, and through iterative optimization, the data points are assigned to the most optimal clusters. However, the limitations of the classical K-means clustering algorithm lie in its computational load and convergence speed. The Quantum Semi-Supervised K -Means Clustering Al algorithm leverages the high-speed properties of quantum computing to quickly calculate the positions of cluster centers in each iteration and accelerates the convergence process through quantum interference effects.
With quantum computing, each step of the K-means clustering calculation can be executed in parallel on a quantum computer, significantly improving the efficiency of the algorithm. Compared to traditional algorithms, the quantum Semi-Supervised K-means Clustering can handle larger datasets within the same amount of computation time and, when dealing with complex data structures, it can offer higher clustering accuracy. First, the quantum matrix product estimation algorithm accelerates the fundamental computations, enabling subsequent learning tasks to be completed in an extremely short time.
Then, the quantum self- training algorithm and the quantum semi-supervised K-means clustering algorithms play roles in classification and clustering tasks, respectively, utilizing the parallelism and efficiency of quantum computing to significantly improve the speed and accuracy of model training and inference. The core of this framework lies in the close integration of quantum computing and classical machine learning methods. During the training process, quantum algorithms are responsible for efficiently processing the data, while classical algorithms perform model optimization based on the data processing.
Through this quantum-classical hybrid approach, WiMi's Quantum Semi-Supervised Learning framework not only fully leverages the advantages of quantum computing but also avoids the limitations of current quantum computing technologies, which are still immature. WiMi leverages quantum supremacy to implement Quantum Semi-Supervised Learning technology, which offers a clear time advantage over classical methods when dealing with large-scale datasets. Through quantum computing, its algorithms are able to complete training in a shorter time and provide more accurate classification and clustering results.
The time complexity of the quantum self-training algorithm and the quantum semi- Supervised K-means clustered algorithm is significantly lower than that of classical algorithms, enabling model training and inference to be completed within an acceptable time frame when handling large-scale data. This advantage makes Quantum Semi-Supervised Learning highly promising in practical applications, especially in scenarios.
