Hybrid computing is the use of both classical and quantum hardware. This includes CPUs, GPUs and quantum processing units (QPUs).
QPUs harness characteristically quantum phenomena, such as superposition and entanglement, to perform certain computations more efficiently than classical computers can. This enables problems to be solved beyond what is possible with solely classical resources.
All QPUs are assisted to some extent by classical computers. Hence, the term hybrid computing is used when the workflow of a quantum-centric algorithm involves a nontrivial amount of classical computation. Many variational quantum algorithms are of this form. In this case, the role of the classical computer is to optimize important parameters in the quantum circuit (e.g., the rotation angles of parameterized quantum logic gates).
TQ42 offers a simple way to access and leverage all these types of hardware, providing business enhancements today and opening the path for quantum utility in the future. TQ42 is connected to the QMware hybrid cloud, where you can select from a suite of high performance devices. Regular benchmarking ensures you can always choose the best tool for the job. For example, virtual (or simulated) qubits often play an important role when prototyping or deploying practical hybrid algorithms.
The tensor network perspective:
Since quantum computers and classical computers have a number of profound differences, it is not straightforward to imagine a form of software that they could both run. An impressive intellectual journey has led to the development of so-called "tensor network" algorithms.
Tensor networks are quantum-inspired, in the sense that they were first developed in order to simulate complex quantum phenomena on classical computers. They are also quantum-compatible because they naturally encompass the circuit model of quantum computing. Most recently, work by Terra Quantum has shown that a key concept in the theory of tensors can be used to decide at which points in an algorithm it is best to switch between quantum and classical processors.
Overall, tensor networks are a crucial aspect of hybrid software, enabling the seamless interfacing of classical and quantum devices.
The quantum machine learning perspective:
Machine learning has been incredibly successful in the past decade or so. The researchers who are developing quantum algorithms therefore have a lot of theoretical tools at their disposal, as well as a great deal of inspiration.
Quantum machine learning, narrowly defined, aims to recreate or surpass the achievements of classical neural networks using quantum circuits. More broadly, hybrid quantum machine learning focusses on how quantum algorithms can complement and enhance these classical architectures.
Consequently, the software being written by hybrid quantum machine learning researchers is an important aspect of hybrid software.
Neural networks & hybrid quantum neural networks
An introduction to neural networks
The optimization perspective:
Just like quantum and classical machine learning can complement and enhance one another, so can quantum and classical optimization. Many practically relevant optimization problems require a heuristic approach, where different methods are combined in order to quickly find a good, yet approximate solution.
Finding heuristics and software methodologies that effectively combine, or hybridize, quantum and classical optimization procedures is an active and emerging field of research.
