See a video clip on NISQ on q-ctrl.com
Recently, I have been reading a lot about the NISQ era. NISQ means Noisy Intermediate Scale Quantum applications. While many quantum platforms already running quantum algorithms with success, the variational quantum algorithms designed for NISQ time, which means " variational quantum algorithms" potentially can improve and increase NISQ devices usefulness.
If I were to bet which industry can use NISQ devices first, I would bet on Quantum Chemistry, combinatorial optimization, machine learning, quantum linear algebra (qiskit) are the obvious ones.
The industries range from drug development in medicine and health science, Logistics, finance automation and optimization, robotics and a few more I left out.
How does it help? Well, let's examine the drug discovery example. There is a VQE algorithm (The Variational Quantum Eigensolver (VQE) is a flagship algorithm for quantum chemistry using near-term quantum computers) that helps potentially find ground state of different molecular Hamiltonian drastically faster, than our best classical super computers. The results potentially leads to understanding of protein folding dynamic simulation, a task that is impossible to run on super computer clusters of today equipped with billions of transistors running on 0 and 1 - the classic sequential calculation concept.
Let's examine an example. You can see the original link here explaining the VGE algorithm.
Example 1:
Using a Ritz variational principle, a quantum computer is trained to prepare the ground state of a given molecule.
"The inputs to the VQE algorithm are a molecular Hamiltonian and a parametrized circuit preparing the quantum state of the molecule. Within VQE, the cost function is defined as the expectation value of the Hamiltonian computed in the trial state. The ground state of the target Hamiltonian is obtained by performing an iterative minimization of the cost function. The optimization is carried out by a classical optimizer which leverages a quantum computer to evaluate the cost function and calculate its gradient at each optimization step.
In this tutorial you will learn how to implement the VQE algorithm in a few lines of code. As an illustrative example, we use it to find the ground state of the hydrogen molecule, H2H2. First, we build the molecular Hamiltonian using a minimal basis set approximation. Next, we design the quantum circuit preparing the trial state of the molecule, and the cost function to evaluate the expectation value of the Hamiltonian. Finally, we select a classical optimizer, initialize the circuit parameters, and run the VQE algorithm using a PennyLane simulator.
PS: more quantum Plugins and ecosystem found here
Above: Example Code input.
Below: The Output example:
The outputs of the function are the Hamiltonian, represented as a linear combination of Pauli operators, and the number of qubits required for the quantum simulations. For this example, we use a minimal basis set to represent the molecular orbitals. In this approximation, we have four spin orbitals, which defines the number of qubits. Furthermore, we use the Jordan-Wigner transformation [2] to perform the fermionic-to-qubit mapping of the Hamiltonian. For a more comprehensive discussion on how to build the Hamiltonian of more complicated molecules, see the tutorial Building molecular Hamiltonians."
See the entire document and additional link here. The above code is credit to pennylane website.
Example 2:
Quantum kernel estimation, variational quantum classifiers are two examples of Supervised and Unsupervised machine learning algorithms and a NISQ device can outshine a classical computer in many cases.
Some of current NISQ devices (Ion Trapping, Superconducting and Photonic) can be used to build quantum circuit Born machines, quantum Boltzmann machines (The main purpose of Boltzmann Machine is to optimize the solution of a problem. It is the work of Boltzmann Machine to optimize the weights and quantity related to that particular problem), and quantum generative adversarial networks.
There are more example to include, but I like to continue the focus on NISQ, Quantum Advantage and touch on Fault Tolerant device and "guess" a reasonable timeline.
As of September 2021 (last time I checked), Google offering a 54 qubit noisy system and claimed "quantum supremacy", running a useless problem solving example, but, to their full credit, it worked awesome. At the time, they "calculated" an equivalent of 10,000 years of classical computing time to solve that particular problem. Since then, Chinese scientists have come close to do the same and classical computers have created advanced algorithms that brought 10,000 years to only 8 years of running on super computers of today!. Well, Google's noisy quantum system solved this useless problem in a couple of hours.
IBM offers 65 qubits, Rigetti works with 31 qubits, ionQ running a 32 qubit system and promising a 64 qubit system sometimes next year (2022).
😆None of these devices (systems) are "good" enough to claim Quantum Advantage -not yet!. They need to run at least with 72 -300 fully algorithmic functionating qubits to win the "advantage" honor and prizes. Imagine 2 to the power of 72, or 2 to the power of 300 is a massive space for parallel computing, which translates to solving "useful" problems in chemistry, global warming, optimization, machine learning, robotics, stock market portfolio optimization, etc. It will change everything.
IBM already announced plans on 1000 qubits by 2023 and a path to 1000,000 superconducting qubits. Google, saying they have a 10 year timeline to a million perfect qubits.
Atom Computing will launch a 100-200 qubit system in 2022, PsiQuantum working on error corrected, fault tolerant systems with one million plus qubits.
Note that currently, a Noise-resilient system not well understood beyond a 100 qubit system and a small set of problems. That makes building a error-free system extremely challenging.
End Post
Comments