Skip to content

Commit d151a8b

Browse files
co9olguyjosh146
co9olguy
and
josh146
authored
Update CHANGELOG for release v0.10 (#687)
* Update CHANGELOG.md * Apply suggestions from code review Co-authored-by: Josh Izaac <[email protected]> * Update .github/CHANGELOG.md * Update CHANGELOG.md * Update .github/CHANGELOG.md * Update CHANGELOG.md * Update .github/CHANGELOG.md Co-authored-by: Josh Izaac <[email protected]> Co-authored-by: Josh Izaac <[email protected]>
1 parent 8863809 commit d151a8b

File tree

1 file changed

+79
-65
lines changed

1 file changed

+79
-65
lines changed

‎.github/CHANGELOG.md

+79-65
Original file line numberDiff line numberDiff line change
@@ -2,9 +2,11 @@
22

33
<h3>New features since last release</h3>
44

5-
* Adds a new device, `default.qubit.tf`, a pure-state qubit simulator written using TensorFlow.
5+
<h4>New and improved simulators</h4>
6+
7+
* Added a new device, `default.qubit.tf`, a pure-state qubit simulator written using TensorFlow.
68
As a result, it supports classical backpropagation as a means to compute the Jacobian. This can
7-
be faster than the parameter-shift rule for analytic quantum gradients
9+
be faster than the parameter-shift rule for computing quantum gradients
810
when the number of parameters to be optimized is large.
911

1012
`default.qubit.tf` is designed to be used with end-to-end classical backpropagation
@@ -34,93 +36,77 @@
3436
[documentation](https://pennylane.ai/en/stable/code/api/pennylane.beta.plugins.DefaultQubitTF.html)
3537
for more details.
3638

37-
* Contains the new template `UCCSD` implementing the Unitary Coupled-Cluster (UCCSD) ansatz
38-
to perform VQE-based quantum chemistry simulations using PennyLane-QChem.
39-
[(#654)](https://github.com/XanaduAI/pennylane/pull/654)
40-
39+
* The [default.tensor plugin](https://github.com/XanaduAI/pennylane/blob/master/pennylane/beta/plugins/default_tensor.py) has been significantly upgraded. It now allows two different
40+
tensor network representations to be used: `"exact"` and `"mps"`. The former uses a
41+
exact factorized representation of quantum states, while the latter uses a matrix product state
42+
representation.
43+
([#572](https://github.com/XanaduAI/pennylane/pull/572))
44+
([#599](https://github.com/XanaduAI/pennylane/pull/599))
45+
46+
<h4>New machine learning functionality and integrations</h4>
47+
4148
* PennyLane QNodes can now be converted into Torch layers, allowing for creation of quantum and
4249
hybrid models using the `torch.nn` API.
4350
[(#588)](https://github.com/XanaduAI/pennylane/pull/588)
4451

4552
A PennyLane QNode can be converted into a `torch.nn` layer using the `qml.qnn.TorchLayer` class:
4653

47-
```python
48-
@qml.qnode(dev)
49-
def qnode(inputs, weights_0, weight_1):
50-
# define the circuit
51-
# ...
54+
```pycon
55+
>>> @qml.qnode(dev)
56+
... def qnode(inputs, weights_0, weight_1):
57+
... # define the circuit
58+
... # ...
5259

53-
weight_shapes = {"weights_0": 3, "weight_1": 1}
54-
qlayer = qml.qnn.TorchLayer(qnode, weight_shapes)
60+
>>> weight_shapes = {"weights_0": 3, "weight_1": 1}
61+
>>> qlayer = qml.qnn.TorchLayer(qnode, weight_shapes)
5562
```
56-
63+
5764
A hybrid model can then be easily constructed:
5865

59-
```python
60-
model = torch.nn.Sequential(qlayer, torch.nn.Linear(2, 2))
66+
```pycon
67+
>>> model = torch.nn.Sequential(qlayer, torch.nn.Linear(2, 2))
6168
```
6269
* Added a new "reversible" differentiation method which can be used in simulators, but not hardware.
70+
6371
The reversible approach is similar to backpropagation, but trades off extra computation for
6472
enhanced memory efficiency. Where backpropagation caches the state tensors at each step during
65-
a forward pass, the reversible method only caches the final pre-measurement state. Compared to
66-
the parameter-shift method, the reversible method can be faster or slower, depending on the
67-
density and location of parametrized gates in a circuit (circuits with higher density of
68-
parametrized gates near the end of the circuit will see a benefit).
73+
a simulated evolution, the reversible method only caches the final pre-measurement state.
74+
75+
Compared to the parameter-shift method, the reversible method can be faster or slower,
76+
depending on the density and location of parametrized gates in a circuit
77+
(circuits with higher density of parametrized gates near the end of the circuit will see a benefit).
6978
[(#670)](https://github.com/XanaduAI/pennylane/pull/670)
79+
80+
```pycon
81+
>>> dev = qml.device("default.qubit", wires=2)
82+
... @qml.qnode(dev, diff_method="reversible")
83+
... def circuit(x):
84+
... qml.RX(x, wires=0)
85+
... qml.RX(x, wires=0)
86+
... qml.CNOT(wires=[0,1])
87+
... return qml.expval(qml.PauliZ(0))
88+
>>> qml.grad(circuit)(0.5)
89+
(array(-0.47942554),)
90+
```
91+
92+
<h4>New templates and cost functions</h4>
7093

71-
* Contains the new template `DoubleExcitationUnitary` implementing the quantum circuit to
72-
exponentiate the Coupled-Cluster double excitation operator. This template is required to
73-
build the Unitary Coupled-Cluster Singles and Doubles (UCCSD) ansatz for VQE simulations.
94+
* Added the new templates `UCCSD`, `SingleExcitationUnitary`, and`DoubleExcitationUnitary`,
95+
which together implement the Unitary Coupled-Cluster Singles and Doubles (UCCSD) ansatz
96+
to perform VQE-based quantum chemistry simulations using PennyLane-QChem.
97+
[(#622)](https://github.com/XanaduAI/pennylane/pull/622)
7498
[(#638)](https://github.com/XanaduAI/pennylane/pull/638)
99+
[(#654)](https://github.com/XanaduAI/pennylane/pull/654)
75100
[(#659)](https://github.com/XanaduAI/pennylane/pull/659)
76-
77-
* Contains the new template `SingleExcitationUnitary` implementing the quantum circuit to
78-
exponentiate the Coupled-Cluster single excitation operator. This template is required to
79-
build the Unitary Coupled-Cluster Singles and Doubles (UCCSD) ansatz for VQE simulations.
80101
[(#622)](https://github.com/XanaduAI/pennylane/pull/622)
81-
[(#659)](https://github.com/XanaduAI/pennylane/pull/659)
82102

83-
* Added module `pennylane.qnn.cost` with class `SquaredErrorLoss`. The module will contain classes
84-
to calculate losses and costs on circuits with trainable parameters.
103+
* Added module `pennylane.qnn.cost` with class `SquaredErrorLoss`. The module contains classes
104+
to calculate losses and cost functions on circuits with trainable parameters.
85105
[(#642)](https://github.com/XanaduAI/pennylane/pull/642)
86106

87-
* The ``default.tensor`` plugin has been significantly upgraded. It now allows two different
88-
tensor network representations to be used: `"exact"` and `"mps"`. The former uses a
89-
exact factorized representation of quantum states, while the latter uses a matrix product state
90-
representation.
91-
([#572](https://github.com/XanaduAI/pennylane/pull/572))
92-
([#599](https://github.com/XanaduAI/pennylane/pull/599))
93-
94107
<h3>Improvements</h3>
95108

96-
* Adds `decomposition` method to PauliX, PauliY, PauliZ, S, T, Hadamard, and PhaseShift gates, which
97-
decomposes each of these gates into rotation gates.
98-
[(#668)](https://github.com/XanaduAI/pennylane/pull/668)
99-
100-
* The `CircuitGraph` class now supports serializing contained circuit operations
101-
and measurement basis rotations to an OpenQASM2.0 script via the new
102-
`CircuitGraph.to_openqasm()` method.
103-
[(#623)](https://github.com/XanaduAI/pennylane/pull/623)
104-
105-
* The QNode Torch interface now inspects QNode positional arguments.
106-
If any argument does not have the attribute `requires_grad=True`, it
107-
is automatically excluded from quantum gradient computations.
108-
[(#652)](https://github.com/XanaduAI/pennylane/pull/652)
109-
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
110-
111-
* The QNode TF interface now inspects QNode positional arguments.
112-
If any argument is not being watched by a `tf.GradientTape()`,
113-
it is automatically excluded from quantum gradient computations.
114-
[(#655)](https://github.com/XanaduAI/pennylane/pull/655)
115-
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
116-
117-
* QNodes have two new public methods: `QNode.set_trainable_args()` and `QNode.get_trainable_args()`.
118-
These are designed to be called by interfaces, to specify to the QNode which of its
119-
input arguments are differentiable. Arguments which are non-differentiable will not be converted
120-
to PennyLane Variable objects within the QNode.
121-
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
122-
123-
* A significant refactor with respect to how QNodes and interfaces mark quantum function
109+
* A significant improvement with respect to how QNodes and interfaces mark quantum function
124110
arguments as differentiable when using Autograd, designed to improve performance and make
125111
QNodes more intuitive.
126112
[(#648)](https://github.com/XanaduAI/pennylane/pull/648)
@@ -141,12 +127,40 @@
141127
The ability to pass `argnum` has been retained for backwards compatibility, and
142128
if present the old behaviour persists.
143129

130+
* The QNode Torch interface now inspects QNode positional arguments.
131+
If any argument does not have the attribute `requires_grad=True`, it
132+
is automatically excluded from quantum gradient computations.
133+
[(#652)](https://github.com/XanaduAI/pennylane/pull/652)
134+
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
135+
136+
* The QNode TF interface now inspects QNode positional arguments.
137+
If any argument is not being watched by a `tf.GradientTape()`,
138+
it is automatically excluded from quantum gradient computations.
139+
[(#655)](https://github.com/XanaduAI/pennylane/pull/655)
140+
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
141+
142+
* QNodes have two new public methods: `QNode.set_trainable_args()` and `QNode.get_trainable_args()`.
143+
These are designed to be called by interfaces, to specify to the QNode which of its
144+
input arguments are differentiable. Arguments which are non-differentiable will not be converted
145+
to PennyLane Variable objects within the QNode.
146+
[(#660)](https://github.com/XanaduAI/pennylane/pull/660)
147+
148+
* Added `decomposition` method to PauliX, PauliY, PauliZ, S, T, Hadamard, and PhaseShift gates, which
149+
decomposes each of these gates into rotation gates.
150+
[(#668)](https://github.com/XanaduAI/pennylane/pull/668)
151+
152+
* The `CircuitGraph` class now supports serializing contained circuit operations
153+
and measurement basis rotations to an OpenQASM2.0 script via the new
154+
`CircuitGraph.to_openqasm()` method.
155+
[(#623)](https://github.com/XanaduAI/pennylane/pull/623)
156+
144157
<h3>Breaking changes</h3>
145158

146159
* Removes support for Python 3.5.
147160
[(#639)](https://github.com/XanaduAI/pennylane/pull/639)
148161

149162
<h3>Documentation</h3>
163+
* Various small typos were fixed.
150164

151165
<h3>Bug fixes</h3>
152166

0 commit comments

Comments
 (0)