Quantum Deep Learning

Quantum Deep Learning

Quantum neural networks, variational circuits, QCNN, hybrid networks, QRNN, autoencoders, QGAN and quantum attention concepts, with a clear distinction between what is mature and what is research.

A quantum neural network and a hybrid model are not the same thing

In a quantum neural network (QNN) all the trainable parameters are gate angles in a quantum circuit, and the output is the expectation value of an observable. A hybrid model combines classical layers (such as Linear and ReLU) with one or more quantum layers, all trained together by backpropagation. The deep quantum models here (QCNN, quantum transformers and others) are early research models, and there is no general evidence of their superiority over classical models on practical problems.

Module lessons

  1. 1Quantum Neural Networks"Quantum neural network" is a common name for a multi-layer parameterised circuit model trained by gradients; it is a l…Advanced
  2. 2Variational Quantum CircuitsA variational quantum algorithm optimises the parameters of a quantum circuit U(θ) in a classical loop to minimise a co…Intermediate
  3. 3Quantum Convolutional Neural Networks (QCNN)The quantum convolutional neural network (QCNN) is a hierarchical variational circuit: convolution layers apply a two-q…AdvancedEarly research model
  4. 4Quantum vs Hybrid Neural NetworksA quantum neural network (QNN) is a variational circuit on its own, in which layers of gates correspond to neural layer…Intermediate
  5. 5Hybrid Quantum–Classical LearningHybrid learning is a loop in which the quantum computer evaluates the model function and its gradients from measurement…Intermediate
  6. 6Quantum Recurrent Neural Networks (QRNN)Quantum recurrent networks (QRNN) are research attempts to process sequences with quantum circuits: a sequence element…Advanced
  7. 7Quantum AutoencodersA quantum autoencoder is a variational circuit that compresses a set of n-qubit quantum states into k < n qubits, by le…AdvancedEarly research model
  8. 8Quantum Generative Adversarial Networks (QGAN)A quantum generative adversarial network (QGAN) is a game between a generator and a discriminator, one or both of which…AdvancedEarly research model
  9. 9Quantum Attention ConceptsQuantum attention is a name covering research proposals to rebuild the attention mechanism and transformer layers with…Advanced
  10. 10Parameterized Quantum CircuitsA parameterised quantum circuit is a circuit containing rotation gates whose angles are real parameters θ; its structur…Intermediate

Experiments in this module

Module quiz

The quiz needs JavaScript.

Local experiment results come from an exact classical state-vector simulation running in your browser, and computation server results come from Qiskit and PennyLane simulators; neither is equivalent to running on a real quantum computer. Memory doubles with each qubit ($2^n$ amplitudes), so the number of qubits is limited to 16 in the browser.