#quantumannealers

[govindhtech]

Convolutional restricted boltzmann machines CRBM Explained

Convolutional restricted Boltzmann Machines

University of California, Berkeley and Lawrence Berkeley National Laboratory researchers revealed a breakthrough in condensed matter physics modelling with industrial partners. They boosted speed by three to five orders of magnitude by meticulously developing a digital hardware accelerator for Convolutional Restricted Boltzmann Machines (CRBM). Simulations may be 100,000 times faster than GPU-based approaches.

The achievement of overcoming a computational barrier opens the door to designing and discovering novel quantum materials with unique properties, such as topological quantum computing potential or high-temperature superconductivity.

The Frustration Computational Challenge

This pioneering study examines geometrically frustrated lattice systems. Highly complicated materials cannot satisfy competing interactions due to their atom arrangement, unlike simple magnetic materials where neighbouring spins align dependably. This “frustration” condition, like three magnets on a triangle opposing each other, causes a huge degeneracy of potential low-energy states. This causes fascinating and often surprising physical occurrences.

Spin liquids have disordered magnetic moments even at absolute zero and behave like quantum fluids rather than solids. Understanding frustrated systems needs accurate simulation, but larger lattices increase computing complexity, making Monte Carlo simulations on CPUs or GPUs impracticable. These systems have so many states that sampling and representation must be rethought.

Innovation in Machine Learning: Variational Wavefunction CRBMs

Berkeley researchers Pratik Brahma, Junghoon Han, Tamzid Razzaque, Saavan Patel, and Sayeef Salahuddin used machine learning to tackle this challenge. RBMs, generative neural networks, were exploited as a powerful variational wavefunction.

A compact and efficient representation of the system's quantum state is learnt by the neural network, focussing on the low-energy states that physicists care about. The RBM shrinks the search space by learning the probability distribution of only the most physically important configurations rather than computing all possible configurations.

A lattice-specific Convolutional Restricted Boltzmann Machine (CRBM) formulation was the breakthrough. Traditional, fully-connected RBMs are inefficient on large lattices because their parameters (connections) quadratically rise with system size. Using the lattice structure's innate translational symmetry, the CRBM overcomes this restriction.

The CRBM uses convolutional filters that match the lattice's unit cell size, like image processing convolutional layers, to find patterns regardless of position. While ensuring that the parameter count is independent of system size, these filters capture localised physical interactions like competing closest and next-nearest neighbour spins. This scaling ability speeds up Monte Carlo sampling and improves the network's representation of difficult states, resulting in faster convergence and more uncorrelated samples.

Custom Silicon for Unmatched Performance

The researchers constructed a CRBM-specific digital hardware accelerator since general-purpose computer hardware slows down even the best algorithm. This custom silicon platform was built using an FPGA, enabling architectural optimisations not feasible in conventional computing.

The accelerator's optimal parallelism architecture could update the spin lattice simultaneously. The surprising speedup is due to optimised bitwise operations, fixed-point weight representations for processing, and most importantly, a hardware design that mirrored the CRBM's convolutional structure to take advantage of translational symmetry.

This complicated hardware-software co-design outperforms variational Monte Carlo techniques on high-end GPUs by three to five orders of magnitude. Important sampling steps might take 33 nanoseconds to 120 milliseconds, depending on the phase repeated.

Validating Exotic Matter Phases

To validate their powerful new tool, the researchers employed the geometrically frustrated Shastry-Sutherland (SS) Ising model, which has a complex and sophisticated phase diagram with long-range ordered fractional plateaus and elusive spin liquid phases.

CRBM hardware simulations of lattices with 324 logical spins were achieved. The simulations, which recovered all known phases of the SS Ising model, proved that the machine could accurately represent and explore frustrated systems' complicated energy landscapes. The customised hardware characterised subtle spin activity at critical locations and inside spin liquid phases, surpassing phase identification. The spin structure factor, which quantifies magnetic order and ties to experimental data like diffuse neutron scattering, confirmed the machine's reliability for basic physics research.

Outperforming Quantum Competitors

Its performance against future quantum technology is one of the most intriguing discoveries. Reports say the CRBM technology is one to two orders of magnitude quicker than cutting-edge quantum annealers, which handle optimisation issues like finding the ground state in frustrated lattices.

However, CRBM hardware outperforms annealers in scalability, room temperature operation, and programmability. Unlike quantum annealers, which require cryogenic temperatures and have inflexible architecture, the FPGA-based CRBM may be easily reprogrammed for different models or simulation conditions. CRBM hardware's efficient integration with a CPU host speeds up variational Monte Carlo computations, paving the way for its rapid adoption in physics labs.

An effective new methodology is established, however it is limited to systems with translational symmetry, restricting its direct applicability to all material classes. By showing that the CRBM hardware can operate as a robust variational wavefunction, this study advances machine learning and specialised digital hardware to solve problems that previously required massive quantum resources.

Future research will expand the CRBM architecture to support more complex symmetries or disordered materials, making it more applicable and making it easier to understand, predict, and find the next generation of quantum materials.

[govindhtech]

Photon-Number Encoding Boosts Quantum-Parallel Computing

Stable Photon-Number Encoding Advances Quantum-Parallel Computation

This breakthrough of using light to encode and process quantum information advances the search for more powerful quantum computers. Researchers Nicholas Chancellor, Uchenna Chukwu, and Mohammad-Ali Miri of Quantum Computing Inc. revealed a method that encodes data in the relative displacement, or photon number, of light modes.

This approach was developed to better defend against quantum system faults. The team uses quantum principles to enable intended parallel processing, going beyond the inadvertent integration of non-Gaussian effects in previous quantum devices. This discovery may lead to more dependable and effective quantum annealers.

Encoding numbers in binary lets computers and digital systems understand, store, and modify them. The simplest representation is bits. Different numbers need different encoding. Popular integer methods like signed magnitude and the two's complement efficiently represent positive and negative numbers and simplify arithmetic. IEEE 754 handles fractional real values.

This method defines a number via a sign (positive or negative), an exponent (which specifies magnitude), and a mantissa (which keeps crucial digits). Number range and accuracy (fraction representation) depend on encoding. Standardised encoding is needed for data transfer and precise calculations across hardware platforms.

Robust Differential Photon Number Encoding Researchers introduced Differential Photon Number Encoding (DiPNE). By storing information using the relative photon number of different light modes, DiPNE can replace phase-based encoding in coherent Ising machines (CIMs).

Optical state displacements provide a practical degree of flexibility for encoding data for quantum parallel computing, supporting this technique. By measuring relative photon number differences, the work shows that information may be reliably stored and processed without accurate phase control, which is often error-prone in quantum computing.

The DiPNE approach uses photon numbers or relative displacement magnitudes in multiple modes. The encoded data is insensitive to squeezing and non-Gaussian changes, preventing frequent errors.

Researchers modelled coherent states passing via a 50:50 beamsplitter. This study showed how input pulse phase inconsistencies impact light dispersion in output channels. Direct homodyne measurement is then utilised to measure encoded data.

Creating High-Fidelity Quantum Superpositions

This encoding method generates high-quality quantum states needed for complex processing. Photon subtraction produces high-quality squeezed state quantum superpositions, according to the study.

This method has more fidelity than current methods that simply yield cat states (superpositions of coherent states). The researchers said mild squeezing and anti-squeezing is not expected to regulate the system's photon count. High-fidelity superpositions—accurate combinations of multiple quantum states—are the goal.

Quantum advantage and explicit non-Gaussian interference Quantum-parallel computation uses quantum mechanics to do several calculations at once. These parallel processing capabilities allow explicit non-Gaussian interference. In contrast to all-optical coherent Ising machines (CIMs), where non-Gaussianity is usually incidental, this intended use is noteworthy.

CIMs are optical systems for difficult optimisation problems. Even though simple CIM implementations are often easily simulatable by classical computers and may not provide a quantum benefit, this study suggests a means to accomplish true quantum processing.

Non-Gaussian states are harder to simulate classically, hence modifying them is necessary to increase computational complexity. This new encoding method can directly leverage non-Gaussianity while preserving a direct encoding of an optimisation problem through interference effects, combining the benefits of CIMs with Gaussian Boson sampling. The study also examines how the Zeno effect can speed up processing faster than typical search methods.

Focussing on Erasure and Control

Since this exploratory study only addresses the encoding component of a wider entropy computing paradigm, the researchers agree that additional work is needed to develop the system.

When designing a full analogue optical quantum optimiser based on DiPNE, consider these future factors:

Need for quantum erasing to account for interference-induced loss channels.

The capacity to selectively fix degrees of freedom not used for encoding without messing up processed quantum information.

Displacement is a good way to encode and process information because it reduces interference from other degrees of freedom. This unique approach is ready for further development because the numerical simulation code is available.

[govindhtech]

D-Wave Quantum Annealers With Landau-Zener Transitions

Landau-Zener Transitions and Domain-Wall Kinetics Explain Quantum Annealer Hysteresis Quantum annealers, which aim to solve difficult problems efficiently, are thought to avoid persistent memory effects seen in conventional computational methods. However, recent experiments have revealed real hysteresis in these quantum systems, a surprising result. This discovery challenges the current understanding of memory creation in physical systems and opens up new avenues for quantum computing research.

Unexpected Memory Effect in Quantum Systems

Hysteresis occurs when a material's reaction, such as magnetisation, lags after the applied external field, demonstrating a form of memory based on earlier exposure. Despite using quantum tunnelling, quantum annealers show magnetisation lagging behind an applied field.

Scientists from Q-MAFIA at Los Alamos National Laboratory, including Frank Barrows, Elijah Pelofske, and Pratik Sathe, and Francesco Caravelli from the Universit~a di Pisa and Scuola Normale Superiore, developed a theoretical framework to explain this unusual behaviour. Their key research indicates that this substantial hysteresis is caused by the complicated interaction between discrete quantum transitions and the continual domain wall movement inside the annealer. In essence, this interaction provides quantum system memory.

Connecting Quantum and Classical Dynamics

The researchers use quantum and conventional physics to explain the observed behaviours. The model incorporates semiclassical domain-wall kinetics and Landau-Zener transitions. Landau-Zener transitions govern system evolution's discrete quantum transitions. Connecting these underlying quantum processes with domain border movement allowed researchers to mimic complex hysteretic activity in investigations. Quantum annealers may imitate and repeat observable behaviour, making them useful for studying programmable quantum hysteresis. This theoretical approach's main strength is its ability to replicate experimental data from numerous annealers and model quantum system behaviour. Importantly, the model helps researchers identify memory effects like temporary negative susceptibilities. This identification proves the annealers are documenting their former status.

Experimental D-Wave Annealers Validation

The study employed D-Wave quantum annealers to simulate magnetic hysteresis in depth to validate their approach. These studies used a quantum annealer with 4,906 qubits to test the model. The ubiquitous Ising model explained the spin-interacting systems under study. An external magnetic field was used to measure the systems' response. On D-Wave quantum annealers, hundreds of qubits were carefully used to form two-dimensional lattices and one-dimensional magnetic chains. Extremely reproducible hysteretic loops were found after carefully adjusting the applied field and measuring the magnetisation. Loops were observed in classical physics contexts where memory effects would not be expected. The simulation findings show that the system's dimensionality and, most crucially, annealing length affect simulation quality. Researchers evaluated how annealing time—the time the D-Wave machine spends finding the lowest energy state—affected results across short and long durations. Two-dimensional simulations yielded smoother and more stable hysteresis loops than one-dimensional simulations, which were more unexpected. The study highlighted how low-dimensional systems' increased noise sensitivity makes simulation harder. Longer annealing times created smoother loops with smaller areas, according to simulations. This makes sense because more time helps the system reach its lowest energy state. The researchers accurately mapped the problem configuration onto D-Wave circuitry using antiferromagnetic gauge transformations. These adjustments made transferring the Ising model to quantum annealing technology practical and efficient.

Relevance and Future Plans

The developed framework accurately predicted how the hysteresis loop will change as the applied field speed varies and replicated complex hysteretic behaviour, including non-monotonic magnetisation reversals and observed variations in the loop's area. The model also replicates local entanglement signs. These findings suggest that programmable annealers are ideal for studying complex, non-equilibrium dynamics in many-body systems. The study establishes these devices as powerful experimental platforms for studying complex quantum phenomena by confirming memory effects. They enable memory-behaving system research. D-Wave quantum annealers can study more complex magnetic systems, which may help us understand magnetism. This study has improved our understanding of quantum memory and its potential uses in quantum technology.

[govindhtech]

How Quantum Annealers Resolve 2D Ising Model Criticality

Quantum annealers are generally employed to determine classical spin systems' low-energy configurations. They address optimisation issues using quantum mechanical notions, unlike universal digital quantum computers.

Their typical operation and distinguishing features:

Adiabatic Evolution: Quantum annealers gradually transform a quantum system from a known initial state, usually a superposition of all possible states, to a final state with the answer. This is adiabatic evolution. A problem-specific Ising Hamiltonian controls its evolution. The term “adiabatic” refers to a slow, steady change in conditions that aims to minimise the goal problem's energy while keeping the system in its ground state.

The D-Wave system implements a transverse-field Ising model with a time-dependent Hamiltonian. A longitudinal energy scale that depicts the classical problem and a transverse field controlled by a function A(s) that encourages quantum fluctuations make up this Hamiltonian. These energy scales evolve according to the annealing schedule parameter, which runs from 0 to 1.

In contrast to ideal zero-temperature quantum computers, quantum annealers use Boltzmann sampling at finite temperatures. The spin configurations at the end of the anneal should follow a Boltzmann distribution if the system stays in contact with its thermal environment (such as the refrigerator or QPU) long enough. This makes them promising for describing complex classical and quantum thermodynamics.

Quantum annealers have received interest for their ability to provide finite-temperature criticality that classical technologies cannot. Their ability to capture classical and quantum phase transitions makes them useful for equilibrium statistical mechanics and finite-temperature criticality.

Despite their potential, quantum annealers have struggled to determine finite-temperature criticality for several reasons:

Noise and Thermal Variations: QPU temperature changes may affect Boltzmann sampling accuracy.

Hardware Limitations: Qubit connectivity, coupling noise, and annealing biases might impair measurement reliability and repeatability.

Critical Slowing Down and Freeze-out: Critical areas incur excessive annealing durations due to system reaction time divergence. A “freeze-out” event can also occur, slowing dynamics and locking the system in metastable configurations.

In Two-Dimensional Ising Model, Quantum Annealers Resolve Finite-Temperature Criticality

A team led by Francesco Campaioli from RMIT University and the Universitá degli Studi di Padova and Gianluca Teza from the Max Planck Institute for the Physics of Complex Systems has improved our understanding of how physical systems change states at realistic temperatures.

In Finite-temperature criticality through quantum annealing, they show how to map and analyse the behaviour of a magnetic material going through a phase transition using a quantum annealer, capturing the entire range of critical behaviour at finite temperatures. This research provides a better knowledge of equilibrium and non-equilibrium processes and opens up new avenues for studying complex systems like magnetism and maybe biological systems.

Critical phenomenon study challenges:

Traditional methods struggle with entanglement, which hinders tensor network approaches in higher dimensions, and critical slowing down, which reduces sampling efficiency.

Even though quantum computing and simulation platforms like digital quantum processors and analogue platforms are promising, the system's response time, QPU temperature fluctuations, and hardware limitations make it difficult to precisely identify finite-temperature criticality with quantum annealers.

Overcoming these obstacles:

The researchers demonstrated that quantum annealers may precisely resolve finite-temperature criticality in complex many-body systems with careful programming.

As a benchmark, they chose the 2D ferromagnetic Ising model, which has a critical point. A novel embedding method reduced edge effects, hardware asymmetries, and magnetic leakage by mapping the simulated system to the annealer's physical design. This requires building physically coupled qubits into logical “superspins” and adding periodic boundary constraints to eliminate edge effects to emulate lattice connection and increase coherence. A calibration approach using local field offsets corrected magnetic leakage biases in superspins with high-amplitude ferromagnetic couplings.

By calibrating the QPU temperature and changing the model's energy scale, they inserted lattices with over 2500 spins. Most of the qubit count was covered by the largest D-Wave Advantage systems. Combined with randomised gauge transformations, this embedding strategy balanced hardware spatial asymmetries.

They ended the anneal at s*=0.5 to keep the system in a dynamic region suitable for thermal equilibrium sampling and avoid freeze-out, which locks the system in metastable configurations.

They also performed in-situ temperature assessments before each batch of samples, a real-time temperature compensation approach not used in previous quantum annealing Boltzmann sampling studies. This allowed consistent finite-temperature sampling across runs and system sizes.

Key discoveries and demonstrations:

The researchers made two vital measurements to demonstrate that they could measure and adjust key solvable model parameters:

After a sudden change, they observed magnetisation throughout a non-equilibrium relaxation and found that a transverse field accelerated convergence to the symmetry-broken ground state. The system often stayed in metastable states and relaxed slowly at small transverse fields.

After starting the system in a metastable state, a fully polarised state opposite the longitudinal field, they assessed the possibility of tunnelling over an energy barrier. The mechanism remained imprisoned at small transverse fields, but the success chance increased drastically as the field extended, suggesting barrier crossing. This underlines how the transverse field helps people escape metastability. At experimental temperature, the mechanism tunnelled through a 60-fold larger barrier than thermal energy. The transverse field levels studied were well below the critical field to ensure that the observed transition was a significant crossing between locally stable configurations rather than a smooth crossover between disordered states.

wider implications: This work establishes a scalable and systematic strategy to finite-temperature criticality research with quantum annealers. Using the great connectivity and coupling tunability of existing annealing architectures, this method can be applied to many classical and quantum systems that can be mapped to the Ising model. These include:

Frustrated magnets

Spinning glasses

Lattice gauge theories

Quantum gravity models

Non-equilibrium quantum thermodynamics.

Abnormal relaxation.

Quantum circuit compilation

The programmable framework can also explore quenched disorder and non-equilibrium dynamics, allowing researchers to study important phenomena that are difficult to study using standard methods. The work finds that accurately calibrated quantum annealers can discover finite-temperature criticality in complex many-body systems, circumventing hardware constraints like the freeze-out point and QPU native topology.

More research on memory effects and meticulous comparisons with conventional algorithms may illuminate the benefits of quantum sampling or optimisation close to criticality.

[govindhtech]

Robotic Inspection from Hybrid Quantum Computing Approaches

Robotic Inspection

D-Wave Quantum Annealers Improve Industrial Optimisation and Promise for Hard Problems

D-Wave Systems' hybrid quantum-classical algorithms and quantum annealers are extending industrial applications, accelerating quantum computing. These technologies are increasingly being researched for their ability to address computationally challenging optimisation problems in complex industrial contexts. Recent investigations show that they can compete with classical techniques for certain issue kinds and are feasible in robotic inspection.

D-Wave's technology is being used to optimise robotic inspection pathways for industrial components with complex three-dimensional geometries from CAD models. The Travelling Salesman Problem (TSP) is reframed in three dimensions. The renowned combinatorial optimisation problem TSP finds the shortest path that stops at exactly one of a defined group of “cities” before returning to the starting point. To alter this problem for robotic inspection, researchers use realistic constraints like open-route requirements and flawed network designs.

D-Wave's quantum-based solvers, which combine classical optimisation approaches with quantum processing principles, have been likened to Google OR-Tools and GUROBI. D-Wave's hybrid solvers (NL-Hybrid and CQM-Hybrid) have produced competitive solution quality while substantially reducing calculation times for car doors, bumpers, toy bears, aircraft parts, and H-spheres in five industry case studies.

Even if Gurobi sometimes produces somewhat better results, D-Wave's solvers' speed advantage is an attractive trade-off, especially in dynamic production contexts where flexibility and quick response are crucial. According to the research, D-Wave solvers provide solutions with AR values between 0.62 and 0.94, indicating near-optimal inspection trajectories. The research measures solution quality using the Approximation Ratio (AR) metric and computing efficiency using the Run Time (RT) metric.

Quantum annealing is a metaheuristic that finds the global minimum of an objective function using quantum fluctuations. Quantum annealers set a system of qubits in an initial ground state and let it grow adiabatically to a final state that encodes the solution, unlike gate-based quantum computers that use controlled qubits in non-binary states. It uses quantum mechanical tunnelling over potential barriers to escape local minima better than classic “hill climbing” methods like Simulated Annealing (SA).

D-Wave's quantum annealers, particularly the Advantage QPU with the Pegasus architecture, are promising quantum processing devices. Pegasus topology, a substantial improvement over previous generations, links each qubit to 15 others. These annealers are not universal quantum computers and have fewer applications than gate-based systems, but they are ideal for optimisation problems, especially those related to the Ising model or QUBO problems.

D-Wave's hybrid solvers are also compared against top-tier classical solvers like CPLEX, Gurobi, and IPOPT on a wider range of optimisation problems. These hybrid solvers address quantum computer scaling limits by dividing problems down and executing parts normally and others on the Quantum Processing Unit (QPU). D-Wave's cloud-based Leap Hybrid Solvers automatically apply constraint penalty terms using quantum hardware and traditional algorithms like Tabu search and simulated annealing.

The hybrid solver of D-Wave provided optimal Binary Linear Programming (BLP) results with linear constraints comparable to CPLEX and Gurobi. However, more variables made it slower to calculate, especially near its scalability limit. At D-Wave's variable limit, adding constraints increased calculation time and decreased solution quality compared to traditional solvers.

D-Wave found poor solutions, often two times the ideal, while addressing integer/binary linear programming problems with quadratic constraints, suggesting it may become stranded in local minima. However, it solved some variable counts faster than IPOPT and CPLEX.

D-Wave's hybrid solver performs best in Binary Quadratic Programming (BQP), which theoretically replicates the Ising Hamiltonian, making these problems ideal for quantum annealers. In this scenario, D-Wave consistently found optimal solutions faster than CPLEX (topped at 1000s runtime) and IPOPT for BQP with 500 binary variables.

Despite longer run times, D-Wave

LeapHybridCQMSolver couldn't compete with Gurobi in computing time and solution quality for Mixed-Integer Linear Programming (MILP) on a power sector Unit Commitment problem. D-Wave may address MILP problems, however its computing advantage appears to be limited to quadratic binary objective functions or limitations.

While Advantage quantum computers and D-Wave's hybrid solvers are starting to compete with traditional optimisation methods, they are only effective for a few problems, especially binary quadratic structures. D-Wave is introducing new topologies like Zephyr and enhancing its gear as the sector grows. Quantum computing could revolutionise industrial optimisation and generate huge profits. Open research data stimulates collaboration and accelerates quantum-inspired solutions.