MicroAlgo Inc. (NASDAQ: MLGO) announced the development of quantum algorithms for feedforward neural networks designed to address performance limitations in traditional neural network training and evaluation, according to a company press release.

The Shenzhen-based company said its quantum algorithm technology builds on feedforward and backpropagation mechanisms while introducing quantum computing capabilities to reduce computational complexity in neural network training.

The technology incorporates three main components: efficient approximation of vector inner products through quantum subroutines, quantum random access memory (QRAM) for storing intermediate values, and natural simulation of regularization effects to reduce overfitting.

MicroAlgo stated that traditional neural network training complexity grows quadratically with the number of neurons and connections, while its quantum approach reduces this to linear complexity. The company said vector inner products are encoded into quantum states using quantum superposition to process multiple dimensions simultaneously.

The QRAM technology allows data storage and access with logarithmic complexity, enabling retrieval of multiple values in a single access, according to the company. MicroAlgo said quantum measurements introduce randomness that helps prevent networks from over-relying on specific weights.

The company indicated potential applications in large-scale data processing for finance and healthcare, real-time decision-making systems for transportation and autonomous driving, and edge computing for Internet of Things devices.

MicroAlgo acknowledged that industrial implementation faces challenges including early-stage quantum computing hardware development, compatibility issues across quantum hardware platforms, and the need for application-specific optimization and testing.

The company develops central processing algorithms and provides software and hardware optimization solutions to customers across various industries.