Neural networks update #5
Conversation
adekusar-drl
requested review from
stefan-woerner,
manoelmarques and
woodsp-ibm
| """Initializes the Circuit Quantum Neural Network. | ||
|
|
||
| Args: | ||
| circuit: The (parametrized) quantum circuit that generates the samples of this network. |
There was a problem hiding this comment.
Parameterized is in () as this optional, ie the circuit is a Circuit and can optionally be parameterized. If so then input_params and weight params needs to be defined, such that the total set of params there match what the circuit needs?
There was a problem hiding this comment.
Now parameterized is not in (). Circuit should be parameterized, otherwise it does not make much sense.
There was a problem hiding this comment.
actually it does make sense, you can use the circuit in a pytorch setting and train a classical NN based on the samples.
| # TODO this should not be necessary... but currently prop grads fail otherwise | ||
| from qiskit import Aer | ||
| self._sampler = CircuitSampler(Aer.get_backend('statevector_simulator'), param_qobj=False) |
There was a problem hiding this comment.
Ok, hopefully we don't need that. If there was any need to do something we should use BasicAer since Aer is still really an optional install and not part of Qiskit core (i.e. Terra)
There was a problem hiding this comment.
Dependency on Aer is removed.
| QiskitMachineLearningError: Invalid parameter values. | ||
| """ | ||
| if num_inputs < 0: | ||
| raise QiskitMachineLearningError('Number of inputs cannot be negative!') |
There was a problem hiding this comment.
Its always nice to have the invalid value etc in these error messages of course.
There was a problem hiding this comment.
Added invalid values to the messages.
| """Initializes the Opflow Quantum Neural Network. | ||
|
|
||
| Args: | ||
| operator: The parametrized operator that represents the neural network. |
There was a problem hiding this comment.
This does not have parameterized in () - but given the following params default to None I guess though this is analgous to the circuit based ine
There was a problem hiding this comment.
Same as for the circuits.
| self.quantum_instance, | ||
| param_qobj=is_aer_provider(self.quantum_instance.backend) | ||
| ) | ||
| # TODO: replace by extended caching in circuit sampler after merged: "caching='all'" |
There was a problem hiding this comment.
FYI This has been merged already Qiskit/qiskit#5934
There was a problem hiding this comment.
Yes, now make use of caching="all".
* moving to public * mypy, spell, other fixes * more on sphinx * more on sphinx * more on sphinx * fix docs * type hints * fix mypy Co-authored-by: Manoel Marques <Manoel.Marques@ibm.com> Co-authored-by: Stefan Woerner <WOR@zurich.ibm.com>
Summary
New implementations of QNNs.
Details and comments
This includes:
Authored by @stefan-woerner, edited by @adekusar-drl.