Leonard Wossnig, CEO of Rahko
Current quantum computers are far from where we need them to be for practical applications due to their high level of ‘noise’ (errors).
If we cannot find a way to use these current and near-term quantum computers, we will need to wait for fully-error-corrected ‘universal’ machines to be developed to see real significant benefit (15-20 years by many estimates).
This is where the software becomes much more than a necessary complement to the hardware. Quantum software has the potential to significantly accelerate our pathway to practically useful quantum computers.
Quantum algorithms
Most quantum algorithms developed to date cannot be run on near-term quantum computers, however there are some that can.
One particular class of algorithm, variational quantum algorithms, is a lead contender for being able to demonstrate near-term quantum advantage.
Variational quantum algorithms
These algorithms allow users to change control parameters of the quantum computer until results match a target property, such as the energy of a molecule – highly relevant to battery manufacturing, room temperature superconductivity, drug discovery and fertilizer manufacturing.
Variational quantum algorithms have already been used to successfully simulate small chemical systems on quantum computers over the last two years, by our team at Rahko and a small handful of teams across the globe.
Chemical Simulation
Broadly speaking, in chemical simulation we look at two types of calculations:
1. Fast, low-cost, low-precision calculations that neglect exact quantum properties
2. High-precision, high-cost calculations
Typically, the first type of calculation is used to filter large pools of candidates, such as candidate drugs. Once a pool has been filtered to a much smaller pool, the second type of calculation is performed to verify exact candidate properties.
This mix allows an optimal use of computational resources.
Quantum computing will likely not directly help with the first type of calculation (low-cost, low-precision), as quantum computing is inherently more expensive and slower.
Machine learning (ML)-based approaches, however, do offer a speedup here. At Rahko, part of our work is in developing classical ML approaches to deliver faster classical solutions for this type of calculation. We can then use quantum computers to generate training data to improve classical ML algorithms.
For the second type of calculation (high-cost, high-precision), quantum computers will bring far greater accuracy at reduced cost. Most importantly, quantum computers will be able to produce accurate simulations where classical methods fail. This will be a game-changing improvement when working with strongly correlated materials, which play a huge role in batteries and room temperature superconductivity.
However, we still face the problem of noise in near-term machines.
There is a solution: quantum machine learning.
Quantum machine learning (‘QML’)
Over the past two years, ML based approaches to running quantum algorithms have borne out powerful results.
Several algorithms have been proposed that, when combined, allow QML approaches to be 10,000,000 times faster than traditional variational quantum algorithms. This means that QML approaches will enable practical gains months, even years, before other variational quantum methods succeed.
Our team at Rahko is working hard to deliver these gains – for the past two years we have developed Hyrax, a QML platform that allows us to rapidly build, test and deploy QML algorithms.
Hyrax relies heavily on variational quantum algorithms and powers all of our state-of-the-art research, helping us to push forward on a QML-enabled pathway to the first commercially valuable practical applications of quantum computing.
With Hyrax, we aim to follow in the footsteps of world leading, UK-born quantum chemistry software, in the tradition of packages such as ONETEP and CASTEP.
The UK quantum future
I strongly believe that QML will play a key role in the UK quantum future.
Investment in QML talent and ventures will give the UK an opportunity to uphold its leading role in quantum chemistry, and a lead role in global quantum computing at large.
This piece was first published as a guest blog post on techUK as part of the techUK Quantum Future Campaign week.
If we cannot find a way to use these current and near-term quantum computers, we will need to wait for fully-error-corrected ‘universal’ machines to be developed to see real significant benefit (15-20 years by many estimates).
This is where the software becomes much more than a necessary complement to the hardware. Quantum software has the potential to significantly accelerate our pathway to practically useful quantum computers.
Quantum algorithms
Most quantum algorithms developed to date cannot be run on near-term quantum computers, however there are some that can.
One particular class of algorithm, variational quantum algorithms, is a lead contender for being able to demonstrate near-term quantum advantage.
Variational quantum algorithms
These algorithms allow users to change control parameters of the quantum computer until results match a target property, such as the energy of a molecule – highly relevant to battery manufacturing, room temperature superconductivity, drug discovery and fertilizer manufacturing.
Variational quantum algorithms have already been used to successfully simulate small chemical systems on quantum computers over the last two years, by our team at Rahko and a small handful of teams across the globe.
Chemical Simulation
Broadly speaking, in chemical simulation we look at two types of calculations:
1. Fast, low-cost, low-precision calculations that neglect exact quantum properties
2. High-precision, high-cost calculations
Typically, the first type of calculation is used to filter large pools of candidates, such as candidate drugs. Once a pool has been filtered to a much smaller pool, the second type of calculation is performed to verify exact candidate properties.
This mix allows an optimal use of computational resources.
Quantum computing will likely not directly help with the first type of calculation (low-cost, low-precision), as quantum computing is inherently more expensive and slower.
Machine learning (ML)-based approaches, however, do offer a speedup here. At Rahko, part of our work is in developing classical ML approaches to deliver faster classical solutions for this type of calculation. We can then use quantum computers to generate training data to improve classical ML algorithms.
For the second type of calculation (high-cost, high-precision), quantum computers will bring far greater accuracy at reduced cost. Most importantly, quantum computers will be able to produce accurate simulations where classical methods fail. This will be a game-changing improvement when working with strongly correlated materials, which play a huge role in batteries and room temperature superconductivity.
However, we still face the problem of noise in near-term machines.
There is a solution: quantum machine learning.
Quantum machine learning (‘QML’)
Over the past two years, ML based approaches to running quantum algorithms have borne out powerful results.
Several algorithms have been proposed that, when combined, allow QML approaches to be 10,000,000 times faster than traditional variational quantum algorithms. This means that QML approaches will enable practical gains months, even years, before other variational quantum methods succeed.
Our team at Rahko is working hard to deliver these gains – for the past two years we have developed Hyrax, a QML platform that allows us to rapidly build, test and deploy QML algorithms.
Hyrax relies heavily on variational quantum algorithms and powers all of our state-of-the-art research, helping us to push forward on a QML-enabled pathway to the first commercially valuable practical applications of quantum computing.
With Hyrax, we aim to follow in the footsteps of world leading, UK-born quantum chemistry software, in the tradition of packages such as ONETEP and CASTEP.
The UK quantum future
I strongly believe that QML will play a key role in the UK quantum future.
Investment in QML talent and ventures will give the UK an opportunity to uphold its leading role in quantum chemistry, and a lead role in global quantum computing at large.
This piece was first published as a guest blog post on techUK as part of the techUK Quantum Future Campaign week.
The French startup ColibriTD has raised €1 million in funding from the Earlybird fund to make quantum computing universally accessible