FAQ
Is your technology a universal (or general) quantum computer?
No, our entropy quantum computer (Dirac) is specially designed to find the ground state of a given Hamiltonian operator. If a problem formulation can be found which takes the form of an Ising model, then for given size constraints, it can be solved with our technology.
How do quantum mechanics come into the operation of EQC?
- Here are papers that describe the building blocks of EQC:
- McCusker 2013 Experimental Demonstration of Interaction-Free All-Optical Switching via the Quantum Zeno Effect
- Chen 2019 Efficient quasi-phase-matched frequency conversion in a lithium niobate racetrack microresonator
Will Dirac devices find an optimal solution every time?
- Dirac devices are capable of finding the ground state of a Hamiltonian operator, but knowing when the ground state has been found is difficult. They will find solutions near the ground state in one of possibly many degenerate states. If the operator describes the desired solution correctly, with all bodies appropriately coded to be in the precision of the device, then the solution will be good in the original problem.
- Especially in the early generations, a trade-off between optimality and time to solution may need to be reached. Sometimes, feasibility will be treated as flexible. Other times, feasibility will be enforced at the expense of optimality. Increased device precision will require less trade-off in these facets.
Where do I find more information on EQC devices?
What makes entropy quantum computing different than quantum annealing?
Comparison of EQC to Quantum Annealing
Quantum Annealing | EQC | |
Quantum Mechanics | Phase transitions under the adiabatic theorem | Quantum Zeno effect with decoherence-free subspaces |
Connectivity | Direct connection physical topology requires embedding to get full connectivity between qubits | Free from topological limitations means full connectivity between qubits without embedding |
Environment | Closed system. The qubits in this environment are kept as free from entropy from heat and electromagnetic radiation as possible. | Open system. The qubits in this environment are expected to interact with the environment, which is engineered into the system. |
Mathematical Domain | Ising Hamiltonian | Hamiltonian on Fock Space |
Shots | Requires thousands of shots to reach statistically significant solution. | The concept of shots is not used. Multiple samples can be taken, but only a small number are required to see meaningful results. |
Where can I find benchmarks against classical or other quantum computing devices?
Not yet, but we are compiling these metrics and will put them into a consumable format.
How many EQC qubits will be required before I can process problems that cannot be solved by classical computers today in a reasonable amount of time?
This is a question with a complex answer. Time efficiency is only one metric that can be used to compare quantum computers to classical. That said, having qubit counts in the six or seven digit range will be required to process linear optimization problems which cannot be solved to a tolerable level of optimality in a reasonable amount of time today. Non-convex problems reach intractability at a number of variables orders of magnitude below the level classical computers can process linear problems at today.
What size of problems can EQC handle?
- We are currently solving problems on the order of 11,000 binary variables.
- We have the capability of solving integer optimization problems on the order of 1,000 variables currently.
How does a user interface with the EQC devices?
- We offer a REST API and a python package which uses the REST API.
- In late 2023, we look to offer a standalone device capable of being installed locally.
What types of problems are appropriate for Dirac-1 vs Dirac-2 and future generations?
- Dirac-1 can accept many more variables than Dirac-2, so a problem with a large number of binary variables may only run on Dirac-1.
- Dirac-2 is capable of representing much larger variable values than Dirac-1, so variables with arbitrarily large magnitudes are best modeled with Dirac-2.
- We may encounter models which do not fit nicely into either device alone, requiring a decomposition approach and using the two devices in conjunction.