Fig. 1: NN architecture implemented on the D-Wave annealer. | Communications Physics

Fig. 1: NN architecture implemented on the D-Wave annealer.

From: Quantum sequel of neural network training

Fig. 1

The input layer consists of 784 neurons, representing pixels of 28 × 28 MNIST images. The hidden and output layers, comprised of 120 and 40 neurons respectively, are encoded as qubits on the quantum annealer. Each neuron in the output layer represents a digit class (0–9), with the 4-fold error-correcting redundancy. An output layer qubit, measured as +1 indicates the corresponding digit class.

Back to article page