Inspiration-Q is a CSIC spin-off that helps forward-looking and innovative companies by delivering new solutions in optimization, simulation and machine learning based on quantum algorithms that work both in quantum as well as in ordinary computers. We make sure you are not only ready for the quantum future, but that your company or firm enjoys quantum advantage now.
Who we are
We are a team of quantum scientists and consultants, with ample experience in the design and simulation of quantum hardware and quantum algorithms. Inspiration-Q is created to bring the best of our research into real-world applications, lowering the costs and barrier to quantum computing and quantum-inspired technologies.
Quantum inspired solutions
In designing quantum algorithms for quantum computers, we learn about what makes those algorithms work and provide computational advantages. We transform this knowledge into quantum inspired solutions, that run in ordinary computers with performance or accuracy benefits, proven by state of the art research.
This means in working with us you get the best of both worlds: quantum algorithms for the quantum computers of the now and the future, and quantum inspired solutions that provide immediate benefit in many real-life production scenarios.
One are of application of quantum inspired solutions is combinatorial optimization. We have developed state-of-the-art optimization solutions that combine classical and quantum-inspired accelerations, and applied them to canonical problems, such as portfolio optimization.
The following plot illustrates a problem with 100 assets from S&P500, each represented with 4 bits in a Markowitz portfolio optimization problem. The space of solutions contains about 1040 portfolios, of which we cannot even extract the right number of valid configurations, due to the enormous size.
As shown in the picture, a good mixture of classical and quantum inspired techniques can outperform simpler classical or quantum solutions. A trivial variational quantum algorithm or a Monte Carlo random sampling can get trapped in local minima or violate constraints (see scattered blue dots on the left). However, a quantum inspired technique finds the frontier of optimal portfolios, with significant improvement in KPI’s such as total return or Sharpe ratios.
Optimizing a portfolio is a static and very limited problem. In real-world scenarios we would like to do predictions over different situations (e.g. changes in volatility, expected returns, etc). In those cases, or in other problem domains such as engineering, we need to do simulations. Our research in this area shows that quantum inspired solutions can provide exponential savings in memory and time.
|Dimension||Resolution||Classical algorithm||Quantum inspired|
|2D||16000||2 Gb||0,000009 Gb|
|2D||200k||254 Tb||0,001 Gb|
|3D||2048||64 Gb||0,001 Gb|
The table above summarizes the gains in memory, and by extension gains in run-time, experienced when modeling a multivariate probability distribution with 2 (2D) or 3 (3D) assets, on a regular discretization. Results come from a Fokker-Planck model for a smooth, differentiable multivariate probability distribution as shown in this [Reference].
These improvements imitate the exponential gains that are obtained when representing those problems in a quantum computer. This is possible because not all hard computational problems exploit all the potential of a quantum computer!
Get quantum now!
Get in touch with us, or sign up to receive news about our products and research outcomes, or follow us on Twitter