Measurement-enabled online quantum processing with amplitude encoding

This paper introduces and experimentally validates a measurement-enabled online quantum reservoir computing protocol that utilizes mid-circuit measurement and reset operations to achieve amplitude encoding, thereby enabling scalable, real-time processing without input buffering or linear runtime overhead.

The Gist

Imagine you have a very complex, chaotic machine—like a swirling storm of water or a tangled ball of yarn—that you want to use to solve puzzles. You can't see inside the machine, but you can poke it with a stick (input) and watch how it reacts (output). This is the basic idea behind Quantum Reservoir Computing (QRC): using the natural, wild movements of a quantum system to process information over time.

For a long time, scientists had a perfect theoretical recipe for how to feed information into this machine, called "amplitude encoding." However, they couldn't actually build it on real quantum computers. The recipe required a magical step: taking a piece of the machine, reading its state, and then instantly "erasing" it to make room for new data, all without stopping the machine's motion. Real quantum computers are fragile; usually, looking at them (measuring them) breaks the magic, and you can't just "erase" a qubit without messing up the rest of the system.

This paper introduces a clever new way to build this machine using real hardware. Here is how they did it, using some everyday analogies:

1. The "Mid-Circuit" Magic Trick

Think of the quantum computer as a long line of people passing a secret note down the chain.

• The Old Problem: To update the note with new information, you had to stop the whole line, read the note, throw it away, and start over. This was slow and broke the "online" flow (processing data as it comes in).
• The New Solution: The authors realized they could use a "mid-circuit measurement." Imagine a person in the middle of the line reading the note, writing down what they saw, and then immediately handing a blank piece of paper to the next person.
• The Reset: In the quantum world, this "handing over a blank paper" is called a reset. By measuring a specific group of qubits (the "input" people) and then resetting them to zero, the math naturally mimics the "erasing" step required by the old theoretical recipe. This allows the machine to keep running smoothly while constantly updating with new data.

2. The "Spy" Technique (Indirect Measurement)

Once the machine is running, the scientists need to know what's happening inside to solve the puzzle.

• The Problem: If you look directly at the "memory" part of the machine (the qubits holding the past data), you disturb it, changing the result. It's like trying to check the temperature of a soup by sticking a thermometer in; the thermometer changes the soup's temperature.
• The Solution: They used ancilla qubits (helper qubits) as "spies."
  • Imagine the memory qubit is a person holding a secret. Instead of asking them directly, you whisper to a "spy" (the ancilla) who is standing next to them. The spy copies a little bit of the secret without the main person even noticing.
  • Then, you measure the spy, not the main person. This gives you the information you need while keeping the main machine's internal state mostly undisturbed.
  • Crucially, they can control how "loud" the spy whispers. They can make the spy whisper very softly (weak measurement) or shout the secret (strong measurement), allowing them to tune how much the machine is disturbed.

3. The Results: A Working Prototype

The team tested this idea on a real quantum computer (an IBM machine with superconducting qubits). They treated the quantum computer like a "black box" reservoir and fed it two types of challenges:

1. Forecasting: Predicting the next number in a chaotic, unpredictable sequence (like trying to guess the next wave in a storm).
2. Memory: Remembering a number from the past and recalling it later (like a short-term memory test).

What they found:

• Their new method worked. The quantum computer successfully processed the data in real-time (online) without needing to pause and save data to a hard drive.
• The results matched perfectly with their computer simulations, proving that the "mid-circuit measurement and reset" trick successfully recreated the theoretical "amplitude encoding" model.
• They could watch the "input" qubits directly and the "memory" qubits indirectly, giving them a full view of the system's behavior.

The Bottom Line

This paper doesn't claim to have built a quantum supercomputer that can cure diseases or predict the stock market tomorrow. Instead, it solved a specific engineering puzzle: How do you make a quantum computer process a stream of data continuously, just like a human brain does, without breaking the delicate quantum state?

They did this by inventing a protocol that uses "measure-and-reset" tricks to clear out old data and "spy" qubits to peek at the results, all while the machine keeps running. This opens the door for scientists to build larger, more complex quantum machines that can learn from time-based data in real life.

Technical Summary

Technical Summary: Measurement-enabled online quantum processing with amplitude encoding

Problem Statement
Quantum reservoir computing (QRC) utilizes the natural dynamics of quantum systems for temporal information processing. The "amplitude encoding" scheme, introduced by Fujii and Nakajima, is a reference model where inputs are injected at each time step by tracing out encoding qubits and preparing a new state parametrized by the input. While theoretically powerful for combining nonlinear codification and state contraction, implementing this scheme on real hardware has remained elusive. The primary challenge is that each input update requires tracing out the state of encoding qubits without interrupting the algorithm's temporal flow, while simultaneously monitoring system observables. Existing approaches often rely on "rewinding" protocols that reconstruct dynamics from sliding time windows, requiring input buffering and making performance dependent on window size and initial states, thus failing to achieve true online operation.

Methodology
The authors propose an experimental online processing design that implements amplitude encoding on quantum hardware using two core mechanisms:

1. Mid-circuit Measurement and Reset for Amplitude Encoding:
   The protocol splits available qubits into "input-reset" qubits and "memory" qubits. At each time step, the input is encoded via a unitary acting on all qubits. Subsequently, the input qubits are measured mid-circuit. Crucially, these qubits are immediately reset to the $|0\rangle$ state. The authors demonstrate that averaging over the stochastic measurement outcomes of these reset operations mathematically introduces the partial trace required by the amplitude encoding map. This allows the reservoir to evolve continuously without halting for input updates.

2. Indirect Measurement for Observable Access:
   To monitor the reservoir's internal dynamics without disrupting the online flow, the authors employ an indirect measurement scheme. Ancillary qubits are entangled with the memory qubits via a controlled interaction (a CNOT gate combined with a rotation gate parameterized by $\alpha$). The ancillae are then measured and reset. This process allows for the extraction of observables from the memory qubits while controlling the measurement backaction (dephasing). The strength of this measurement is tunable, ranging from weak measurements to projective limits, and is mathematically equivalent to continuous-variable ancilla measurements previously defined in spin systems.

The theoretical framework models the system as a sequence of completely positive and trace-preserving (CPTP) maps, where the monitored state evolves along a quantum trajectory determined by measurement outcomes. The protocol is validated through numerical simulations comparing the proposed circuit model against the original spin model and through a proof-of-principle experiment on the IBM "basquecountry" quantum processor (Heron r2, 156 qubits).

Key Contributions

• Online Protocol Realization: The paper introduces a hardware-native protocol that realizes amplitude encoding as a dynamic circuit, preserving online operation and avoiding the need for input buffering.
• Partial Trace via Reset: It identifies mid-circuit measurement followed by reset as the specific hardware primitive that implements the partial trace dynamics underlying amplitude encoding.
• Tunable Monitoring: The scheme enables the simultaneous monitoring of the entire system (input and memory qubits) by isolating the internal reservoir dynamics into a separate channel via indirect measurements, allowing for the study of measurement backaction.
• Hardware Implementation: The authors provide a proof-of-principle implementation on superconducting qubits, demonstrating the feasibility of the approach on current noisy intermediate-scale quantum (NISQ) devices.

Results
The authors evaluated the protocol on two standard benchmark tasks:

1. Santa Fe Task: Forecasting a chaotic time series.
2. Short-Term Memory (STM) Task: Storing and retrieving past information.

• Simulation vs. Theory: Numerical simulations using Qiskit density-matrix backends showed perfect agreement with the theoretical spin model dynamics for both input qubits (under projective measurement) and memory qubits (under indirect measurement).
• Hardware Performance: Experiments on the IBM hardware demonstrated results close to the ideal spin model and circuit simulations.
  • For the forecasting task, the hardware results closely matched the simulation, with minor gaps attributed to Trotterization errors and finite shot noise rather than hardware noise.
  • For the STM task, differences appeared at higher memory depths ($\eta \approx 5$), where hardware noise played a more significant role.
• Optimization: The measurement strength parameter ($g$) was optimized for each task, with values of $g=2.05$ for forecasting and $g=0.256$ for STM, consistent with previous findings on optimal measurement strength for these tasks.

Significance
The paper claims to provide a practical route toward scalable hardware implementations of amplitude-encoded quantum reservoir computing. By successfully combining amplitude encoding with tunable measurement backaction in a single online design, the work enables systematic experimental studies of complex quantum reservoirs as monitored dynamical systems. The authors emphasize that this approach opens the door to observing the full system evolution while isolating the internal reservoir dynamics, a capability previously difficult to achieve on real hardware. The work does not claim to solve the noise limitations of current hardware but rather establishes a functional framework for online QRC that can be refined as hardware improves.