Optimization

qnetti.optimize(cost, settings, step_size=0.1, num_steps=20, verbose=False, init_opt_dict=None, filepath='./', filename='', step_only=False, **meta_opt_kwargs)

Minimizes the cost function and find the optimal settings.

The objective of the optimization is \(\min_{\vec{\theta}} Cost(\vec{\theta})\). In the optimization a gradient descent algorithm is applied. In each iteration of gradient, descent, the settings \(\vec{\theta}`\) are updated as

\[\vec{\theta}' = \vec{\theta} - \eta \nabla_{\vec{\theta}} Cost(\vec{\theta})\]

where \(\eta\) is the step size and \(\nabla_{\vec{\theta}}Cost(\vec{\theta})\) is the gradient of the cost function evaluated as \(\vec{\theta}\).

Parameters:

• cost (function) – The function to be minimized. This function will be called as cost(settings).

• settings (qml.numpy.ndarray) – A PennyLane Numpy tensor that has the requires_grad=True attribute.

• step_size (float) – The step to take in the direction of steepest descent. Default step_size=0.1.

• num_steps (int) – The number of iterations of gradient descent to perform. Default num_steps=20.

• verbose (bool) – If True, the iteration step and cost will be printed every 5 iterations.

• init_opt_dict (dictionary) – An optimization dictionary to serve as a warm start for the optimization. The optimization will continue from the provided dictionary.

• step_only (bool) – If True, the cost is not evaluated. Default False.

• meta_opt_kwargs (keyword arguments) – Generic keyword arguments to write the optimization dictionary.

Returns:

A dictionary of optimization data.

Return type:

dictionary

The return dictionary has the following keys: * "settings_list" - list[list[float]], The settings considered in each optimization step. * "cost_vals" - list[float], The cost evaluated in each optimization step. * "min_cost" - float, The smallest cost evaluated during optimization. * "opt_settings" - list[float], The optimal settings to achieve the minimum cost. * "step_size" - float, The step size used in the optimization. * "datetime" - string, The date/time at which the data is optimization begins. * "opt_step_time" - float, The time needed to make each optimization step. * "error" - boolean, Indicates whether or not an error occurred during optimization.

qnetti.qubit_probs_qnode_fn(prep_node, meas_wires=None, dev_kwargs={}, qnode_kwargs={})

Constructs a qnode function that returns the probabilities for local qubit measurements. The resulting quantum circuit prepares the state described by the prep_node and then applies arbitrary qubit unitaries to each wire before measurement.

Parameters:

• prep_node (qnetvo.PrepareNode) – An ansatz for the network state preparation.

• meas_wires (list[int]) – The wires to apply arbitrary qubit measurements to. If None, all qubits are measured.

• dev_kwargs – Keyword arguments to pass to the PennyLane device constructor. Useful keys are include "name", which specifies the device (default: "default.qubit"), and "shots", which specifies the integert number of shots to run during circuit execution (default: None).

• qnode_kwargs (dict) – Keyword arguments passed through to the PennyLane qnode decorator.

Returns:

A qnode function that returns the qubit probabilities for the circuit and the device that evaluates the qnode. The function is called as qnode(settings) where len(settings) == 3 * num_wires.

Return type:

tuple(qml.QNode, qml.Device)