Quantum analog-encoding for correlated Gaussian vectors and their exponentiation with application to rough volatility

This paper proposes new quantum algorithms for the efficient encoding of correlated Gaussian vectors and their exponentiations into quantum states, demonstrating a potential quantum advantage for simulating rough volatility models used in finance.

The Gist

Imagine you are a professional chef trying to recreate a very complex, secret sauce that changes its flavor slightly every single time you make it. This "sauce" represents the unpredictable movement of stock prices in the financial markets.

To make this sauce perfectly, you need to follow a recipe that involves thousands of tiny, interconnected ingredients (data points). In the world of finance, these ingredients are "correlated Gaussian vectors"—basically, a massive web of random numbers that are all linked together. If one ingredient moves, the others react in a specific, predictable way.

The Problem: The "Kitchen Nightmare"

Currently, if a bank wants to simulate these complex market movements to predict risks, they use classical computers. However, as the number of ingredients (the complexity of the market) grows, the classical computer hits a wall. It’s like trying to stir a pot that is getting bigger and bigger every second; eventually, the chef can't stir fast enough, and the whole process slows to a crawl. This is known as the "cubic complexity" problem—every time you double the ingredients, the work doesn't just double; it increases eightfold!

The Solution: The "Quantum Super-Stirrer"

This paper proposes a new way to do this using Quantum Computing. Instead of a chef with a spoon, imagine a "Quantum Super-Stirrer" that can manipulate all the ingredients simultaneously through a magical property called amplitude encoding.

Here is how the researchers' "recipe" works:

1. The Perfect Mix (Quantum Analog-Encoding)
Instead of writing down every single ingredient on a piece of paper (which takes too much space), the researchers found a way to "encode" the entire web of ingredients directly into the vibrations (amplitudes) of a quantum state. It’s like instead of listing every grain of salt in a soup, you just capture the "essence" of the saltiness in the steam.

2. The "Magic Multiplier" (Exponentiation)
In finance, we don't just care about the ingredients; we care about how they multiply each other (this is called "exponentiation"). This is used to model "Rough Volatility"—the jagged, "rough" way that market prices actually jump around in real life. The researchers created a mathematical trick to perform this multiplication inside the quantum computer without breaking the delicate quantum state.

3. The "Smart Scale" (Quantum Amplitude Estimation)
Once the quantum computer has finished its magical stirring, you can't just look at it to see the result (because looking at a quantum state destroys it!). The researchers developed a "Smart Scale" (called QAE) that allows us to peek at the results and get a very accurate measurement of the "total flavor" (the statistical properties) without ruining the whole batch.

Why does this matter? (The "Quantum Advantage")

The researchers did the math to see if this is actually faster than the old way. They looked at "Rough Bergomi" models—the gold standard for modern financial modeling.

They discovered that for certain types of market movements, their quantum method is significantly faster than even the best classical computers. While a classical computer might get stuck in a "kitchen nightmare" of endless stirring, the quantum approach stays efficient, even as the market becomes more complex and "rough."

In Short:

This paper provides the foundational blueprints for a quantum financial simulator. It moves us from a world where we struggle to simulate complex, jagged market movements to a future where we can model them with precision and speed, helping to predict financial storms before they hit.

Technical Summary

Technical Summary: Quantum Analog-Encoding for Correlated Gaussian Vectors and Their Exponentiation

Authors: Tassa Thaksakronwong and Koichi Miyamoto
Date: April 27, 2026
Field: Quantum Computing / Quantum Finance

---

1. The Problem: Exact Simulation of Rough Volatility

In mathematical finance, realistic modeling of market phenomena (like the "volatility smile") requires rough volatility models, such as the rough Bergomi model. These models are driven by fractional stochastic processes (e.g., fractional Brownian motion) that are "rougher" than standard Brownian motion.

Mathematically, these models rely on the exponentiation of auto-correlated Gaussian processes. Classically, simulating a Gaussian vector $\vec{x} \sim \mathcal{N}(0, \Sigma)$ with a dense covariance matrix $\Sigma$ requires $O(N^3)$ time for Cholesky decomposition and $O(N^2)$ for sampling. While approximation methods exist, they often lack universality or exactness.

Existing quantum approaches for fractional processes (like those using the Quantum Fourier Transform) are often restricted to uniform grids, impose boundary constraints (e.g., $Y_0 = Y_T = 0$), or produce approximations rather than exact simulations. There has been a lack of a framework for the exact, structure-free analog encoding of exponentiated Gaussian processes.

2. Methodology: Quantum Analog Encoding and Non-linear Transformation

The authors propose a quantum algorithmic framework to prepare and manipulate quantum states that encode these processes in their amplitudes (analog encoding).

• State Preparation of $|x\rangle$: They develop an algorithm to prepare the normalized state $|x\rangle = \vec{x}/\|\vec{x}\|$, where $\vec{x}$ is a realization of a correlated Gaussian vector. This is achieved using block-encoding of the covariance matrix $\Sigma$ and Quantum Amplitude Amplification (QAA).
• Exponentiation via QSVT: To handle the non-linear requirement of exponentiation ($e^{\vec{x}}$), the authors utilize Quantum Singular Value Transformation (QSVT).
• Overcoming Complexity Bottlenecks: A major technical hurdle in previous non-linear transformation methods was that the complexity scaled exponentially with the norm of the vector ($\|\vec{x}\|$). The authors solve this by observing that while $\|\vec{x}\|_2$ grows with $N$, the maximum component $\|\vec{x}\|_\infty$ remains bounded by a constant $\Xi$ with high probability. They construct a specialized polynomial that performs the exponential mapping on a shrinking interval around zero, keeping the transformation cost $O(1)$ relative to $N$.
• Cumulative Sums: They provide a method to simulate the "noise" (increments) of a process and use a cumulative sum matrix ($L_N$) to recover the path, which can improve the condition number of the covariance matrix.

3. Key Contributions

1. First Exact Framework: The first quantum algorithmic framework for the exact, structure-free analog encoding of exponentiated fractional Gaussian processes.
2. End-to-End Resource Analysis: A complete complexity analysis from state preparation to the extraction of classical statistics (like the time integral of variance) via Quantum Amplitude Estimation (QAE).
3. Robustness to Structure: Unlike previous methods, this approach does not require the covariance matrix to have a specific structure (like Toeplitz or low-rank), making it universally applicable to dense matrices.
4. Numerical Scaling Analysis: A detailed study of how the complexity scales with the dimension $N$ for specific financial processes (Riemann-Liouville fBM, standard fBM, and fractional Ornstein-Uhlenbeck).

4. Results and Complexity

The gate-depth complexity for preparing the normalized state $|x\rangle$ is:
$$\tilde{O}\left( \frac{\|\Sigma\|_F}{\lambda_{\max}} \kappa^{1.5} \right)$$
where $\|\Sigma\|_F$ is the Frobenius norm, $\lambda_{\max}$ is the maximal eigenvalue, and $\kappa$ is the condition number.

• Quantum Advantage:
  • In the well-conditioned case ($\kappa = O(1)$), the complexity is $\tilde{O}(\sqrt{N})$, providing a sextic speedup over the classical $O(N^3)$ Cholesky method.
  • For fractional Gaussian noises, the complexity is sub-linear to the dimension $N$, potentially outperforming classical FFT-based methods.
• Rough Bergomi Application: For the rough Bergomi variance process, the overall gate-depth complexity ranges from $\tilde{O}(N^2)$ to $\tilde{O}(N^3)$. While this is a "mild" advantage over classical methods, it provides a clear path toward quantum advantage as error tolerances ($\epsilon$) become more stringent.

5. Significance

This paper establishes the foundational primitives for quantum-enhanced financial modeling. By providing a way to exactly encode and exponentiate rough volatility paths, it bridges the gap between abstract quantum algorithms and practical, high-fidelity financial simulations. It moves the field beyond constant-volatility Black-Scholes models toward the complex, "rough" stochastic volatility models used in modern quantitative finance.