MISIS: Protocol for Quantum Computing

The main articles are:

• Quantum Computers and Quantum Communications
• Machine Learning

2025: Optimizing Quantum Algorithms

Researchers at MISIS University have developed a protocol for quantum computing that improves the search for optimal solutions. The approach is based on the targeted launch of special noise channels. In the future, the development will increase the accuracy and speed of calculations. The university announced this on September 24, 2025.

Quantum machine learning problems, despite their great potential, face serious training and optimization difficulties. Due to many possible solutions, not all of which are optimal, the algorithm can "get stuck" without reaching the best solutions. The protocol developed by scientists of NUST MISIS will allow to regulate optimization landscapes using special noise channels.

Typically, noise interferes with the efficient operation of quantum algorithms. Any interaction with the environment - random fluctuations in temperature or electromagnetic fields - lead to errors in calculations. Experts have demonstrated that the use of special noise channels significantly smooths out small-scale fluctuations in the loss function and allows for better solutions.

When we train a model, whether it's a classical neural network or a quantum algorithm, it has a loss function. This is a measure of how wrong her approach to solving the problem is: the higher the loss, the worse. There can be many model parameters, for example, rotation, phase, weight, etc. Each combination of these parameters gives its own result and the loss function assigns a number to this result - "height." Imagine: you stand on a mountain and try to get down to the lowest point. The height indicates how far you are from the target. There are many small pits and depressions on the way and you can easily get stuck in them without reaching the goal. Usually this happens - we wander and fall into local traps. Our method is as if the pits were covered with sand. It fills small hollows, levelling the surface, and the path becomes easier: We no longer linger and can move on. Thus, the addition of noise - regularization - smooths out the landscape and greatly simplifies the search for an optimal solution, - said Nikita Nemkov, Ph.D., senior researcher at the Laboratory of Quantum Information Technologies at NITU MISIS.

The protocol provides for the creation of a certain amount of noise for certain elements in a quantum scheme. As a result, the loss function is smoothed out. Scientists have tested the algorithm on test problems and a quantum convolution neural network. In both cases, the protocol improved the result: the chance to find the right solution turned out to be several times higher compared to traditional approaches.

The difficulty of learning variation quantum algorithms and quantum machine learning models is well known. The protocol we proposed can be combined with the existing method of mitigating local minima - a quantum optimizer of the natural gradient, and can also complement the set of methods for optimizing quantum loss functions. The technical implementation of the protocol does not require a large amount of additional resources and can be used both in classical quantum circuit simulators and on real quantum devices, "said Alexei Fedorov, director of the Institute of Physics and Quantum Engineering at NITU MISIS, head of the scientific group of the RCC" Quantum Information Technologies. "