For perspective, Diane asked me about a month ago to explain quantum computing. My answer was pretty simple, in standard computing, there is a 1 and a 0. So there are two states. In quantum computing there is a 0, 1, and another 0 and 1. So the possible combination of 4 items is 16. so Quantum can vae 16 states, standard computing can have 2, so quantum computing is 8 times more powerful! A bit of an oversimplification if I do say so myself!
Want to solve business problems really fast? Quantum Computing (using quantum mechanics, the science of how tiny particles like electrons behave, to process information in ways regular computers can’t) could change how you handle tough tasks. This post explains quantum computing and examples of how it can be used for businesses based on a video by Gurobi Optimization. Submitted Diane Hart Alexander, MBA, MHA, ALEX Chair and CEO.
Key Points
- What is Quantum Computing?
- Quantum computers (machines that use quantum mechanics to compute differently than regular computers) use qubits (quantum bits, the basic units of quantum information that can be 0, 1, or both at once) instead of regular bits (basic units of regular computing, either 0 or 1).
- Quantum annealers (special quantum computers designed to solve certain optimization problems) work on special problems.
- They might solve hard problems faster, but they’re still being worked on.
- What are Qubits and Why Are They Cool?
- Qubits can be in superposition (a qubit’s ability to be 0, 1, or both at the same time), trying lots of answers at once.
- They can be entangled (when qubits are linked so one changes the other instantly), making some tasks quicker.
- Why They’re Cool: Qubits help quantum annealers solve problems like QUBO (Quadratic Unconstrained Binary Optimization, a problem with only 0 or 1 variables and a squared-number goal, no extra rules) by showing variables (unknowns in a problem, like amounts or choices) in a smart way. But too few qubits and weak connectivity (how qubits link in a machine) make big problems hard.
- Optimization Problems
- Optimization problems (tasks to find the best answer, like making the most money or saving time) range from simple to complex.
- Linear programs (problems with simple goals and rules, like maximizing profit) are easy.
- Quadratic programs (problems with squared numbers in the goal, adding complexity) and nonlinear programs (problems with complex, non-straight-line rules) are trickier.
- Integer programs (problems using only whole numbers) are tough.
- Non-convex problems (super complicated problems with messy solution spaces) are very hard.
- Quantum computing wants to solve non-convex problems faster.
- Quantum Annealers and QUBO
- Quantum annealers solve QUBO.
- Binary variables (variables that are either 0 or 1) are used in QUBO.
- Quadratic (using squared numbers, like x² or xy, in the goal) defines the goal in QUBO.
- Other problems can turn into QUBO by adding penalty terms (extra numbers to enforce rules).
- Example: Add a penalty to stop two variables from both being 1.
- Turning Problems into QUBO
- Constraints (rules solutions must follow, like variables adding to 1) like “sum of variables equals 1” become penalties, needing lots of qubits.
- Inequality constraints (rules like “less than or equal to”) need extra variables using unary expansion (showing a number as a sum of 0s and 1s) or binary expansion (showing a number with binary digits, using fewer variables).
- Big constraints make “fully connected” problems (where every variable connects to every other), needing tons of qubits, which is hard for today’s machines.
- Challenges with Quantum Annealers
- D-Wave’s (a company making quantum annealers) machines have ~5,000 qubits but can only handle ~150 variables because of connectivity limits.
- Embedding (fitting a problem into a machine’s setup) uses extra qubits, leaving less room for big problems.
- Tools like Gurobi often do better than quantum annealers right now.
- Examples
- Maximum Cut Problem (splitting a graph’s points into two groups to have the most connections between them): Fits QUBO well.
- Maximum Independent Set (finding graph points with no connections between them): Uses penalties to avoid connected points.
- Gurobi solved a 58-point independent set in 0.02 seconds, showing it’s really strong.
- Quantum vs. Regular Solvers
- Quantum annealers are heuristic (giving okay but not always perfect answers), unlike Gurobi, which finds the best answers.
- Regular solvers are better with continuous variables (numbers that can be decimals) and tricky constraints.
- Quantum machines aren’t very exact (precision – accuracy in calculations) and have connectivity issues, so they struggle with big problems.
- Hybrid approaches (mixing quantum and regular methods) might help later.
- What’s Next?
- Machines like D-Wave’s Pegasus graph (a setup for qubits with better connections) and Zephyr graph (an even better-connected setup) are improving.
- Quantum annealers could team up with regular solvers in hybrid systems for fast first guesses.
- In 5–10 years, quantum computing might do more, but regular solvers are best now.
Institutions
- Gurobi Optimization: A company that makes tools for solving optimization problems, hosting the talk.
- D-Wave: A company that builds quantum annealers, like Advantage (a D-Wave quantum computer with ~5,000 qubits).
- Fujitsu and Hitachi: Use regular FPGA (field-programmable gate array, customizable hardware) machines for optimization, not quantum.
Call to Action
Curious about optimization? Watch the Gurobi tech talk on YouTube for more on quantum annealers. Want to discuss this article or anything Quantum/AI?
Text me “quantum” to 713.918.9951 and I will immediately respond.
For your convenience key definitions below:
Definitions
- Advantage: A D-Wave quantum computer with ~5,000 qubits.
- Binary expansion: Showing a number with binary digits, using fewer variables.
- Binary variables: Variables that are either 0 or 1.
- Bits: Basic units of regular computing, either 0 or 1.
- Connectivity: How qubits link in a machine.
- Constraints: Rules solutions must follow, like variables adding to 1.
- Continuous variables: Numbers that can be decimals.
- Embedding: Fitting a problem into a machine’s setup.
- Entangled: When qubits are linked so one changes the other instantly.
- FPGA: Field-programmable gate array, customizable hardware.
- Fully connected: When every variable connects to every other.
- Gurobi: Software for solving optimization problems.
- Heuristic: Giving okay but not always perfect answers.
- Inequality constraints: Rules like “less than or equal to.”
- Integer programs: Problems using only whole numbers.
- Linear programs: Problems with simple goals and rules, like maximizing profit.
- Logistics: Managing supply chains.
- Maximum Cut Problem: Splitting a graph’s points into two groups to have the most connections between them.
- Maximum Independent Set: Finding graph points with no connections between them.
- Non-convex problems: Super complicated problems with messy solution spaces.
- Nonlinear programs: Problems with complex, non-straight-line rules.
- Optimization problems: Tasks to find the best answer, like making the most money or saving time.
- Penalty terms: Extra numbers to enforce rules.
- Pegasus graph: A setup for qubits with better connections.
- Precision: Accuracy in calculations.
- Quadratic: Using squared numbers, like x² or xy, in the goal.
- Quadratic programs: Problems with squared numbers in the goal, adding complexity.
- Quantum annealers: Special quantum computers designed to solve certain optimization problems.
- Quantum Computing: Using quantum mechanics, the science of how tiny particles like electrons behave, to process information in ways regular computers can’t.
- Qubits: Quantum bits, the basic units of quantum information that can be 0, 1, or both at once.
- QUBO: Quadratic Unconstrained Binary Optimization, a problem with only 0 or 1 variables and a squared-number goal, no extra rules.
- Scheduling: Planning tasks efficiently.
- Superposition: A qubit’s ability to be 0, 1, or both at the same time.
- Unary expansion: Showing a number as a sum of 0s and 1s.
- Variables: Unknowns in a problem, like amounts or choices.
- Zephyr graph: An even better-connected setup for qubits.
Discover more from alex.realestate & alex ai
Subscribe to get the latest posts sent to your email.