Quantum Computing News

Noisy quantum circuit modeling is revolutionized by the groundbreaking TUSQ simulator, which achieves exponential speedups.

From materials science to medicine, quantum computing could transform many fields. The restricted and expensive availability of real quantum hardware, which frequently has lengthy wait times, is still a major obstacle. This calls for the creation of complex simulators that can replicate the operation of quantum circuits on actual, noisy quantum hardware with high accuracy and scalability.

You can also read Quantum Information Science And Engineerings Drug Discovery

To tackle this crucial issue, University of Chicago scholars Siddharth Dangwal, Tina Oberoi, and Ajay Sailopal, together with their associates, have introduced TUSQ – Tracking, Uncomputation, and Sampling for Noisy Quantum Simulation. By cleverly eliminating pointless computations and reusing computational resources, this innovative method significantly speeds up simulations of noisy quantum circuits, exhibiting previously unheard-of speedups over current techniques.

When it comes to noise, traditional quantum circuit simulation (QCS) encounters a dilemma. Noisy QCS introduces stochastic operations, but noiseless QCS can be effectively carried out utilizing State Vector Simulation (SVS) by multiplying a quantum state vector with deterministic unitary matrices.

A naive SVS approach necessitates executing the complete series of matrix-vector multiplications anew for each sample in order to appropriately account for these probabilistic noise effects. This results in a significant S-fold time overhead, where S is the number of samples. Density Matrix Simulation (DMS) is a solution that uses matrices to represent quantum states and matrix-matrix multiplications to account for noise in a single circuit execution. However, DMS is not viable for a large number of qubits due to its quadruple memory overhead when compared to SVS.

Similar to popular simulators like CUDA-Q and Qiskit Statevector Simulator, TUSQ uses several SVSs to simulate noise, which keeps the memory footprint four times smaller than DMS. However, it also addresses the time overhead that results. The Error Characterization Module (ECM) and the Tree-based Execution Module (TEM), two cutting-edge modules, are responsible for TUSQ’s exceptional efficiency. These modules’ basic idea is to find and remove calculations that are redundant or comparatively irrelevant in order to provide much quicker noisy QCS.

You can also read Quantum Circuit Complexity Reveals Hidden Quantum Phases

Error Characterization Module (ECM): Streamlining Circuit Executions

The main objective of the ECM is to analyze the many circuits produced by stochastic noise channels and find examples that result in the same output, hence lowering the overall number of distinct circuit executions. Two crucial steps are used to do this:

• Error Realization (ER) Tallying: A noisy quantum circuit can be thought of as a classical average of several circuits, each with “fixed noisy gates” that are sampled from the stochastic channels. This is known as error realization (ER) tallying. The frequency of these “error realizations” (ERs) is monitored by TUSQ. TUSQ simulates the related circuit only once and samples its output state vector s times if a specific ER happens s times. This is a much more cost-effective method than doing s separate simulations. This works especially well because ERs with low Hamming weights are far more common in modern quantum computers, which generally have low error rates.
• ER Commutation: This technique goes beyond simple counting to find situations in which various ERs can nevertheless provide the same output state vector. Using a set of commutation rules for Pauli gates and CNOTs, TUSQ achieves this by “pushing” noisy gates as far to the right in the circuit as it can without changing the noiseless circuit. The number of unique circuits that need to be simulated is further decreased if, following this procedure, two different ERs produce similar new ERs by combining their respective shot counts.

You can also read Nokia Quantum Computing Vision For Digital Communication

Tree-based Execution Module (TEM): Reusing Computation

By taking advantage of chances for computational reuse, the TEM optimizes the execution of the minimal set of circuits with unique outputs that the ECM has produced. This module includes:

• Depth-first Tree Traversal (DFTT): TUSQ uses a tree structure to depict the different circuits, with nodes standing in for state vectors at different locations and edges for gates. Similar to classical computer architecture, TUSQ employs a rollback-recovery technique rather than precalculating each circuit from scratch. By using the inverse of gates, it “uncomputes” the final state vector for a single circuit (a leaf node) before continuing along a different branch to calculate the state vector for a new circuit. This drastically cuts down on unnecessary matrix-vector multiplications. Importantly, TUSQ assumes that all gates, even noisy ones, are unitary, which has been verified for a wide range of realistic noise models, including depolarizing and measurement noise, and even decoherence using Pauli-twirling approximations. This makes this computation possible. TUSQ maintains a small memory footprint and allows DFTT to be parallelized utilizing available memory because it does not save intermediate states, in contrast to other simulators that use memorizing techniques. By decreasing operations from a naive implementation to, where |E| is the number of edges and b is the number of possibilities for a noisy channel, DFTT provides an asymptotic advantage.
• Pruning: To increase speed even more, TUSQ finds “insignificant circuits” that appear seldom after ECM because they have a negligible impact on the distribution of the final output. By removing these branches from the tree, computation time is reduced. This is essential for efficiency even if it creates a tiny, controlled disturbance. TUSQ samples a subset of insignificant circuits when their combined contribution is considerable, adjusting their probabilities to maintain the overall output distribution contribution. With user-defined hyperparameters, the average relative fidelity difference introduced by this trimming can be adjusted from 2.1% to a maximum of 8.7%.

You can also read Matrix Product States MPS Quantum Computing Simulation

Unprecedented Performance and Impact

The performance improvements brought about by TUSQ’s optimizations are astounding. Using an Nvidia A100 GPU, the simulator was tested on 186 benchmarks, including QAOA, Adder, Bit Code, Phase Code, and GHZ circuits.

Highlights of the performance include:

• A 12.53x speedup over CUDA-Q and a 52.5x speedup over Qiskit on average.
• The average speedup rises to 55.42x over Qiskit and 23.03x over CUDA-Q for larger benchmarks (above 15 qubits).
• TUSQ occasionally outperformed Qiskit by up to 7878.03x and CUDA-Q by up to 439.38x.
• In less than 820 seconds (819.87 seconds), TUSQ successfully simulated a 30-qubit Adder circuit, which would normally take more than 10 hours on similar simulators.
• TUSQ showed significant improvements over the newly proposed TQSim simulator, attaining an average speedup of 68.6x and increase to 493.4x quicker. This demonstrates the algorithmic advantages of TUSQ, since TQSim depends on more memory and memoization, while TUSQ’s speedup is intrinsic to its depth-first traversal and uncomputation methodology.

You can also read HodgeRank in Quantum Topological Signal Processing at SUTD

While TUSQ’s speedup tends to grow with the number of qubits, it may decrease significantly for deeper circuits or greater error rates because more branches become “significant” and pruning becomes less effective. TUSQ’s anticipated 1% error rate, however, is typical of existing systems, and future technology advancements will only increase TUSQ’s viability.

The creation of TUSQ, a potent and effective instrument for modeling noisy quantum circuits, represents a major advancement in the study of quantum computing. TUSQ speeds up the transition to scalable quantum processing by allowing researchers to investigate bigger and more intricate quantum algorithms through resource reuse and clever computational overhead management. There are plans to offer an open-source version of TUSQ, which should further spur innovation in the area.

You can also read Noise Model Verified On IQM’s Superconducting Quantum Chip