Quantum algorithm for solving high-dimensional linear stochastic differential equations via amplitude encoding of the noise term

This paper proposes two quantum algorithms—one based on the Dyson series and the other on the Euler-Maruyama method—to solve high-dimensional linear stochastic differential equations by using amplitude encoding for the state and a quantum pseudorandom number generator to encode the noise term, achieving a polylogarithmic scaling in the dimension $N$.

The Gist

Imagine you are a weather forecaster trying to predict the path of a massive, swirling hurricane. This hurricane isn't just one single point moving across a map; it is a complex, high-dimensional system of wind speeds, air pressures, and temperatures across thousands of different locations, all being buffeted by unpredictable, chaotic gusts of wind.

In mathematics, this is a High-Dimensional Stochastic Differential Equation (SDE).

• "High-Dimensional" means there are thousands of variables to track at once.
• "Stochastic" means there is randomness (the unpredictable wind gusts) involved.

Currently, even our best supercomputers struggle with this. If you try to track every single tiny gust of wind in a massive system, the math becomes so heavy that the computer eventually "chokes."

This paper proposes a new way to solve this problem using Quantum Computers, and it does so by changing how we "write down" the information.

---

1. The Problem: The "Library" vs. The "Ghost"

To solve these equations, a computer needs to keep track of the state of the system (the hurricane's current status).

• The Old Way (Binary Encoding): Imagine you have a library with 1,000 books, and each book represents one variable (like wind speed in one city). To tell the computer the state of the hurricane, you have to physically walk to every single book and write down a number. If the hurricane has a million variables, you’ll be walking forever. This is slow and inefficient.
• The Quantum Way (Amplitude Encoding): Instead of a library of books, imagine the hurricane is a ghost. The ghost doesn't occupy specific books; instead, its "presence" is spread out across the entire room. By adjusting the "thickness" or "intensity" of the ghost in different spots, you can represent all 1,000 variables at once using a single quantum state. This is called Amplitude Encoding. It is incredibly compact—you can represent a massive amount of data using very few "quantum bits" (qubits).

2. The Big Hurdle: The "Chaos" Problem

The paper identifies a massive problem: How do you encode "randomness" into a ghost?

In a normal equation, you can tell a computer, "The wind is always 10mph." That’s easy to encode. But in an SDE, the wind is random. It’s like trying to tell a ghost to be "randomly thick" in a way that follows the laws of physics. If you just throw random numbers at a quantum computer, you break the delicate quantum state.

3. The Solution: The "Digital Dice" (PRNG)

The author solves this by using a Quantum Pseudorandom Number Generator (PRNG).

Think of it like this: Instead of trying to capture "true" chaos from the universe (which is hard), we use a very complex, high-speed "digital dice roller" built directly into the quantum circuit. This dice roller is so sophisticated that it looks and acts like true randomness, but because it is a mathematical circuit, the quantum computer can use it to shape the "ghost" (the amplitude encoding) perfectly.

4. The Two Recipes: Dyson vs. Euler

The paper offers two different "recipes" (algorithms) to cook this mathematical soup:

• Recipe 1: The Dyson Series Method (The Perfectionist): This method tries to follow the exact, smooth laws of physics. It’s like trying to trace the path of a bird by calculating every microscopic muscle twitch. It is incredibly accurate but requires the "wind" (the noise) to be well-behaved and "full-rank" (meaning the randomness hits every part of the system).
• Recipe 2: The Euler-Maruyama Method (The Practical Traveler): This method is more like a hiker taking steps. Instead of calculating every tiny twitch, it says, "I'll look at where I am, take a step, see how the wind blows, and then take another step." It’s slightly less precise, but it’s much more robust—it works even if the randomness is messy or only affects certain parts of the system.

Why does this matter?

By using these quantum "ghost" methods, the paper proves that we can achieve an exponential speed-up.

In plain English: A problem that might take a classical supercomputer centuries to solve because it has too many variables could, in theory, be solved by this quantum algorithm in minutes. This could revolutionize how we model everything from the stock market and chemical reactions to the way galaxies form in the early universe.

Technical Summary

Technical Summary: Quantum Algorithm for Solving High-Dimensional Linear Stochastic Differential Equations via Amplitude Encoding of the Noise Term

Author: Koichi Miyamoto
Date: April 28, 2026

---

1. Problem Statement

The paper addresses the challenge of solving high-dimensional Linear Stochastic Differential Equations (SDEs) of the form:
$$dX_t = A(t)X_t dt + B(t)dW_t$$
where $X_t \in \mathbb{R}^N$ is a high-dimensional random process and $W_t$ is an $m$-dimensional Brownian motion.

While quantum algorithms for Ordinary Differential Equations (ODEs) exist, extending them to SDEs is non-trivial due to the noise term ($B(t)dW_t$). Most existing quantum SDE approaches use binary encoding (representing numbers via bit strings), which scales linearly with the dimension $N$. To achieve an exponential speed-up in $N$, the author aims to use amplitude encoding, where the vector $X_t$ is stored in the amplitudes of a quantum state. The primary technical hurdle is how to amplitude-encode the stochastic noise term, which cannot be treated as a simple deterministic function of an index.

2. Methodology

The author proposes two distinct quantum algorithmic frameworks to generate the "history state"—a quantum state that simultaneously encodes the values of the process $X_t$ at multiple time points $\{t_0, t_1, \dots, t_r\}$.

A. Dyson Series-Based Method (Exact Approach)
This method is designed for cases where the diffusion matrix $B(t)B(t)^T$ is full-rank.

• Core Idea: It utilizes the exact recurrence relation $X_{n+1} = \Phi_n X_n + \Delta_n$, where $\Phi_n$ is the time-evolution operator and $\Delta_n$ is the noise increment.
• Noise Encoding: The author introduces a novel way to encode the noise by utilizing a Quantum Pseudorandom Number Generator (PRNG) circuit. By combining a PRNG with arithmetic circuits, the algorithm constructs a state that encodes the realizations of the noise in its amplitudes.
• Implementation: The time-evolution operator $\Phi_n$ is approximated using a truncated Dyson series and implemented via block-encoding. The resulting linear system is solved using a Quantum Linear Systems Solver (QLSS) to prepare the history state.

B. Euler-Maruyama (EM)-Based Method (Robust Approach)
This method is designed for cases where $B(t)B(t)^T$ is rank-deficient (i.e., the noise dimension $m < N$), a common scenario in real-world applications.

• Core Idea: It uses the standard Euler-Maruyama discretization: $X_{n+1} \approx X_n + A(t_n)X_n \Delta t + B(t_n)\Delta W_n$.
• Implementation: Instead of the complex Dyson series, it constructs the block-encoding of the EM transition matrix directly from the block-encodings of $A$ and $B$. While this introduces discretization error, it is more computationally efficient and handles low-rank noise terms naturally.

C. Estimation of Expectations
Beyond state preparation, the paper provides methods to extract physical quantities (expectations of functions of $X_t$) using overlap estimation. This allows for the calculation of:

• Multi-time functions: Quantities depending on $X_t$ at various time points.
• Terminal-time functions: Quantities depending only on the final state $X_T$.

3. Key Contributions

1. Amplitude Encoding of Noise: The most significant contribution is the method to encode random variables into quantum amplitudes using PRNG circuits, bypassing the limitations of function-loading methods.
2. Two-Tiered Algorithmic Approach: Providing both a high-precision Dyson-based method and a robust, rank-deficient-friendly EM-based method.
3. History State Generation: Developing a way to prepare a single quantum state representing the entire trajectory of a high-dimensional stochastic process.
4. Complexity Analysis: Rigorous proof that the algorithms achieve an exponential speed-up in terms of the dimension $N$ (scaling as $\text{polylog}(N)$).

4. Results and Complexity

• Speed-up: The algorithms require only $\text{polylog}(N)$ queries to the underlying oracles (for $A$ and $B$), providing an exponential advantage over classical Monte Carlo methods that scale linearly with $N$.
• Dyson Method Complexity: The query complexity to the controlled $A$ oracle is roughly $\tilde{O}\left( \frac{\kappa_{BBT} \alpha_A T}{\max\{1, \eta T\}} \right)^{d+1} \left( \frac{\|x_0\|^2 + N\sigma^2 T}{\epsilon} \right)^{d/2}$, where $d$ is the degree of the function being estimated.
• EM Method Complexity: The EM method offers a more streamlined complexity, particularly advantageous for terminal-time functions, scaling with the discretization error and the condition number of the system.

5. Significance

This work bridges the gap between quantum ODE solvers and the practical requirements of stochastic modeling. By enabling the simulation of high-dimensional SDEs, this research has profound implications for:

• Quantitative Finance: Pricing complex derivatives and modeling market volatility in high-dimensional asset spaces.
• Quantum Chemistry/Physics: Simulating chemical reactions and cosmic inflation where randomness is inherent.
• Large-scale Numerical Computation: Providing a blueprint for how quantum computers can handle "noisy" high-dimensional data more efficiently than classical supercomputers.