Breaking the Barrier: Engineered Dissipation to Overcome Barren Plateaus in Quantum Algorithms

A team of researchers from the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC) has unveiled a groundbreaking approach to tackle one of the most significant challenges in quantum computing: barren plateaus. This phenomenon severely restricts the performance of variational quantum algorithms (VQAs), which are crucial for solving optimization problems on noisy quantum computers.

In their study, entitled "Engineered Dissipation to Mitigate Barren Plateaus", authors demonstrate that incorporating properly engineered Markovian losses after each layer of a unitary quantum circuit can enhance the trainability of quantum models. This finding challenges the conventional belief that dissipation is detrimental to quantum technologies.

Barren plateaus are a phenomenon that occurs in variational quantum algorithms, which are an important class of algorithms used in quantum computing. These plateaus represent a significant challenge because when a quantum algorithm enters a barren plateau, it becomes extremely difficult to improve its performance, similar to being stuck on flat terrain without knowing which direction to move to find the lowest point. Also, they worsen with size, limiting the scalability of quantum algorithms.

Barren plateaus leads to significant inefficiencies in training. The researchers identify the necessary forms of dissipation processes and establish that optimizing these processes is efficient. Their work benchmarks the general applicability of this strategy through both synthetic and practical quantum chemistry examples, showcasing its effectiveness across various domains.

The researchers suggest using a technique called "engineered dissipation". This involves intentionally introducing a controlled form of noise or interaction with the environment after each layer of the quantum circuit. Unlike uncontrolled noise, this engineered dissipation is designed to help the algorithm learn better and avoid getting stuck in barren plateaus. The team emphasizes that while noisy quantum platforms are often seen as a limitation, their findings suggest that non-unitary operations can significantly improve the accuracy of calculations.

This advancement marks a step towards overcoming the current limitations in quantum computing, paving the way for enhanced applications in quantum chemistry, machine learning, and beyond. The research not only highlights the potential of engineered dissipation but also opens new avenues for further exploration in open-system dynamics and quantum algorithms.

Sannia, A., Tacchino, F., Tavernelli, I. et al. Engineered dissipation to mitigate barren plateaus. npj Quantum Inf 10, 81 (2024). https://doi.org/10.1038/s41534-024-00875-0

Photo: Illustrative diagram of the optimized cost function landscape for a system using an engineered dissipation approach.