The problem: connecting concepts, mathematics and computation
Quantum computing brings together several kinds of reasoning. A state has a mathematical representation, an operation has a physical interpretation, and a measurement produces classical data. Understanding a computation requires following the relationships between all three.
Terms such as superposition, interference and entanglement become useful when learners can define them, calculate their consequences and recognise the conditions under which they apply. Analogies can introduce a concept, but they need to lead to a precise account that supports prediction and explanation.
The educational challenge is to connect these foundations with practical work. A learner should be able to explain why a circuit produces a particular distribution, which assumptions make an algorithm valid and how a change in the implementation could affect the result. These abilities provide a basis for assessing unfamiliar methods as the field develops.
An axiomatic foundation
An axiomatic formulation makes the rules of the computational model explicit. For the finite-dimensional systems used in the circuit model, the starting point is a complex Hilbert space: a vector space equipped with an inner product. The formulation then specifies valid states, allowed transformations, composition and measurement.
| Concept | Mathematical and physical meaning |
|---|---|
| States | A pure state is represented by a normalised vector, with vectors differing only by a global phase representing the same physical state. More generally, density operators are positive semidefinite and have unit trace; they include pure and mixed states. |
| Composition | The Hilbert space of a composite system is the tensor product of its constituent Hilbert spaces. This permits entangled states as well as product states. |
| Evolution | The evolution of a closed system between measurements is described by unitary transformations. These preserve the inner product and the normalisation of state vectors. |
| Measurement | The Born rule connects a state and a chosen measurement to the probabilities of classical outcomes. A full measurement description also specifies the state update when the quantum system is retained. |
| General processes | Quantum channels describe deterministic processes, including noise, through completely positive, trace-preserving linear maps. Unitary evolution is a special case. |
These definitions give learners a common language for checking an argument. They also separate the ideal circuit model from the additional descriptions needed for mixed states, noise and interactions with an environment. A course can introduce that generality progressively while stating the scope of each model.
Building the mathematical understanding
Complex numbers, linear algebra and probability provide the working language. Inner products, basis changes, eigenvectors and tensor products should be connected to the quantum concepts they express. Learners need opportunities to work through these relationships, with mathematical preparation matched to the material being introduced.
For a pure state, amplitudes are complex numbers whose squared magnitudes give probabilities in a specified measurement basis. Relative phase can affect interference and later measurement statistics, while a global phase leaves the physical state unchanged. Working through a small circuit makes the distinction calculable and gives the terminology a precise meaning.
The aim is to develop enough fluency to move between a verbal description, a state representation, a matrix calculation and a circuit diagram. IBM Quantum Learning’s mathematical introduction to quantum information illustrates this progression from individual systems to composite systems and circuits.
From principles to algorithms and implementation
An algorithm introduces further questions: how the problem is encoded, which operations are available, what the measurement reveals and how the output is interpreted. Learners should be able to connect each of these choices to the task the algorithm is intended to solve.
Claims about computational advantage require their assumptions and resource model to be stated. Relevant costs can include input preparation, oracle access, circuit size, repeated measurements and classical processing. Education should prepare learners to examine a comparison with classical methods and distinguish a theoretical result from evidence obtained in a particular experiment.
Implementation then connects the model to software and hardware. A useful exercise asks learners to predict a result, construct the circuit, run it and interpret the observations. Finite sampling, numerical approximations and device noise need separate consideration when explaining a difference between a prediction and measured data.
Preparing students, researchers and developers
Students need a structured progression that develops mathematical confidence alongside conceptual understanding. They should practise deriving results, explaining assumptions and applying definitions before moving to more advanced algorithms and applications.
Researchers need to formulate questions precisely, identify the limits of a model and assess the evidence behind a claim. This includes reading technical literature, choosing appropriate baselines and deciding whether a proposed use of quantum computation is justified for the problem.
Developers need to translate a mathematical specification into a program while retaining its intended meaning. They should understand the role of SDK conventions, compilation, execution settings and statistical output well enough to inspect a workflow and explain its behaviour.
The depth and emphasis differ, but all three groups benefit from the same foundation: a precise account of the computational paradigm and the ability to connect that account to the work they perform.
Learning through derivation, practice and feedback
Rigorous learning should combine mathematical exercises with implementation. Derivations make reasoning visible; programming tasks test whether the learner can carry it into a working representation. Explanations of the resulting data reveal whether the calculation and the experiment have been understood together.
Assessment should examine both the answer and the reasoning behind it. Feedback can address an incorrect assumption, a missing normalisation, a mistaken basis convention or an unsupported interpretation of a result. Progress then means being able to solve and explain a new problem with increasing independence.
QuantumClass
QuantumClass is QuantGates’ education platform, which aims to address this need through structured learning, practical programming and feedback. The approach connects conceptual foundations with the skills needed to develop and evaluate quantum computations.
Its published learning offering includes an IBM quantum programming course with Python and Qiskit exercises, graded assignments and instructor feedback. A research programme extends the focus to assessing quantum fit, formulating a problem and developing a testable quantum or hybrid workflow.
The goal is to help learners build the understanding needed to reason about quantum computation, assess its suitability and continue into research or development with a sound foundation. The QuantumClass learning page describes the available courses and research programme.