SHENZHEN, China, April 24, 2026 /PRNewswire/ — MicroAlgo Inc. (the “Company” or “MicroAlgo”) (NASDAQ: MLGO), today announced that they have developed a set of quantum algorithms for feedforward neural networks, breaking through the performance bottlenecks of traditional neural networks in training and evaluation. This innovative quantum algorithm is based on the classic feedforward and backpropagation algorithms, leveraging the powerful computational capabilities of quantum computing to greatly enhance the efficiency of network training and evaluation, and it brings a natural resistance to overfitting.
The feedforward neural network is the core architecture of deep learning, widely applied in fields such as image classification, natural language processing, and speech recognition. However, traditional neural network algorithms face challenges such as high computational overhead, high risk of overfitting, and long training times when dealing with large-scale data and complex models. Quantum computing, with its potential for exponential acceleration, provides a brand-new pathway to address these issues.
Specifically, quantum computing can significantly reduce computational complexity in training neural networks by efficiently handling large-scale matrix and inner product operations. Meanwhile, the unique data storage and retrieval methods of quantum computing can efficiently manage intermediate values during the training process, greatly improving training efficiency and resource utilization. These characteristics make quantum algorithms an ideal choice for enhancing neural network performance.
The quantum algorithm technology developed by MicroAlgo this time is based on the classic feedforward and backpropagation mechanisms, optimizing key computational steps by introducing efficient quantum subroutines.
First, the efficient approximation of vector inner products. The key to neural network training lies in weight updates, and weight updates are inseparable from the computation of inner products between vectors. In traditional methods, the complexity of computing inner products grows quadratically with the number of neurons and connections, resulting in low computational efficiency. MicroAlgo’s quantum algorithm technology introduces quantum subroutines based on the principles of quantum state superposition and interference, which can robustly approximate vector inner products while significantly reducing computational complexity. Specifically, input vectors are encoded into quantum states, utilizing quantum superposition to process computations across multiple dimensions simultaneously. Subsequently, approximate results are extracted through quantum measurements, with a complexity that is only linearly related to the number of neurons, breaking through the limitations of classical methods.
Second, the introduction of quantum random access memory. In neural network training, a large number of intermediate values (such as activation values and error values) need to be stored and quickly retrieved in subsequent stages. Traditional storage methods not only consume significant storage resources but may also lead to inefficient data retrieval. To address this, MicroAlgo’s algorithm utilizes quantum random access memory (QRAM) technology to implicitly store intermediate values in quantum states. QRAM allows data to be stored and accessed with logarithmic complexity, making the training process more efficient. Additionally, due to the superposition property of quantum states, QRAM can retrieve multiple values simultaneously in a single access, further accelerating the training process.
Furthermore, the natural simulation of regularization effects. Overfitting is a common problem faced by neural networks, typically mitigated by adding regularization terms or using techniques such as random dropout. MicroAlgo’s quantum algorithm, due to its unique quantum state characteristics, can naturally mimic the effects of regularization techniques during the training process. For example, there is a certain degree of randomness in quantum measurements, which helps prevent the network from overly relying on specific weights. Additionally, the probabilistic distribution characteristics of quantum computing make weight updates more diverse, thereby enhancing the model’s generalization ability.
The training time of traditional neural networks typically grows exponentially with the increase in network size, whereas this quantum algorithm reduces the training time complexity to a linear level. This improvement is mainly attributed to: the efficient approximate computation of vector inner products significantly reducing computational overhead; the fast storage and retrieval of QRAM avoiding redundant computations; and the parallel computing capability of quantum superposition states accelerating the processing of batch data.
Although quantum algorithms themselves have absolute advantages in certain applications, the principles and logic they propose can also provide new ideas for classical algorithms. For example, by introducing concepts such as approximate inner products and random storage, classical heuristic algorithms with effects similar to quantum algorithms can be designed. Although these algorithms have higher complexity, they still hold practical value in certain specific scenarios.
The development of this quantum algorithm by MicroAlgo has opened new prospects for the enterprise application of quantum machine learning. First, in large-scale data processing, such as in the fields of finance and healthcare, the demand for large-scale data processing is growing rapidly. This quantum algorithm, through its efficient inner product computation and data management capabilities, can quickly analyze and process large-scale data, providing support for areas such as financial risk assessment and genomic research.
In real-time decision-making systems, such as intelligent transportation and autonomous driving, real-time decision-making systems need to rapidly process large amounts of sensor data and respond accordingly. The efficiency and robustness of this algorithm make it an ideal choice for supporting such systems.
Additionally, in the fields of edge computing and the Internet of Things, with the proliferation of IoT devices, edge computing is gradually becoming mainstream. The lightweight design and efficient computational characteristics of this quantum algorithm make it suitable for resource-constrained edge devices, contributing to the construction of an intelligent IoT ecosystem. In the future, this quantum algorithm can also serve as a bridge for the integration of quantum and classical computing, further promoting the popularization of machine learning technologies by optimizing the performance of classical algorithms.
Of course, although MicroAlgo’s quantum algorithm demonstrates immense potential, its industrial implementation still faces some challenges. For example: the development of quantum computing hardware is still in its early stages, and achieving large-scale quantum computing requires overcoming technical bottlenecks; the compatibility and portability issues of quantum algorithms necessitate the development of solutions adaptable to various quantum hardware platforms; optimization and debugging for specific application scenarios still require extensive research and experimentation.
This time, the quantum algorithm developed by MicroAlgo not only marks a significant leap in the performance of feedforward neural networks but also opens a new chapter in the integration of quantum computing and artificial intelligence. Through breakthroughs in computational efficiency, resource utilization, and model generalization capabilities, this algorithm provides new ideas for addressing key challenges in the field of deep learning. In the future, with the continuous improvement of quantum computing hardware and software ecosystems, this technology is expected to drive the implementation of more innovative applications.
This breakthrough technology showcases the potential of interdisciplinary collaboration, bringing together the intellectual achievements of quantum computing, machine learning, and optimization algorithms. It not only expands the application boundaries of quantum algorithms but also provides new inspiration for the optimization of traditional algorithms. Particularly in fields such as real-time decision-making, edge computing, and the Internet of Things, its impact will be even more profound. The successful development of MicroAlgo’s quantum algorithm is not only a technical achievement but also a prelude to artificial intelligence entering the era of quantum computing. In the future, we look forward to the further development of this technology, bringing unprecedented value to more industries and scenarios.
About MicroAlgo Inc.
MicroAlgo Inc. (the “MicroAlgo”), a Cayman Islands exempted company, is dedicated to the development and application of bespoke central processing algorithms. MicroAlgo provides comprehensive solutions to customers by integrating central processing algorithms with software or hardware, or both, thereby helping them to increase the number of customers, improve end-user satisfaction, achieve direct cost savings, reduce power consumption, and achieve technical goals. The range of MicroAlgo’s services includes algorithm optimization, accelerating computing power without the need for hardware upgrades, lightweight data processing, and data intelligence services. MicroAlgo’s ability to efficiently deliver software and hardware optimization to customers through bespoke central processing algorithms serves as a driving force for MicroAlgo’s long-term development.
Forward-Looking Statements
This press release contains statements that may constitute “forward-looking statements.” Forward-looking statements are subject to numerous conditions, many of which are beyond the control of MicroAlgo, including those set forth in the Risk Factors section of MicroAlgo’s periodic reports on Forms 10-K and 8-K filed with the SEC. Copies are available on the SEC’s website, www.sec.gov. Words such as “expect,” “estimate,” “project,” “budget,” “forecast,” “anticipate,” “intend,” “plan,” “may,” “will,” “could,” “should,” “believes,” “predicts,” “potential,” “continue,” and similar expressions are intended to identify such forward-looking statements. These forward-looking statements include, without limitation, MicroAlgo’s expectations with respect to future performance and anticipated financial impacts of the business transaction.
MicroAlgo undertakes no obligation to update these statements for revisions or changes after the date of this release, except as may be required by law.
