Adds single qubit decomposition functions #52

purva-thakre · 2021-06-11T18:36:18Z

WIP for 4 single qubit decomposition functions.

To do before marking PR ready for review :

Check format --> black
Check style --> PEP8

tests/test_general_decompositions.py

src/qutip_qip/decompositions/general_decompositions.py

purva-thakre · 2021-06-13T23:27:34Z

src/qutip_qip/decompositions/general_decompositions.py

+    except ZeroDivisionError:
+        input_array = input_array


Discuss at June 14 meeting

Don't think this exception has to be added. If this function's goal is to check if the matrix is unitary or not then a matrix whose determinant is zero is not a unitary matrix.

Yeah you can return False already.

However, I'm not sure if normalizing is needed. It seems very unlikely to me that a not normalized gate will be unitary after the normalization. If someone provides something with |det| != 1, he/she already missed the point of providing a unitary.

Or is there a special reason?

purva-thakre · 2021-06-13T23:28:52Z

src/qutip_qip/decompositions/general_decompositions.py

+        except ZeroDivisionError:
+            input_array = input_array


Discuss at June 14 meeting

Same as above. If the determinant is zero then do not need to make this exception. det(U) = 1

src/qutip_qip/decompositions/general_decompositions.py

BoxiLi · 2021-06-14T11:28:22Z

src/qutip_qip/decompositions/general_decompositions.py

+
+import numpy as np
+
+def normalize_matrix(input_array)-> np.array:


It would be nice if we can have an optional parameter in the main function that allows turning off some checks if the decomposition is used e.g. for qutip predefined gates. Those gates should, by construction, be valid.

src/qutip_qip/decompositions/general_decompositions.py

BoxiLi · 2021-06-14T12:00:52Z

src/qutip_qip/decompositions/general_decompositions.py

+    except ZeroDivisionError:
+        input_array = input_array


Yeah you can return False already.

However, I'm not sure if normalizing is needed. It seems very unlikely to me that a not normalized gate will be unitary after the normalization. If someone provides something with |det| != 1, he/she already missed the point of providing a unitary.

Or is there a special reason?

BoxiLi · 2021-06-14T12:48:00Z

src/qutip_qip/decompositions/verify_decompositions.py

+#    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+###############################################################################
+
+# add functions to check the calculated decomposition is equivalent to input array. 


If you mean checking whether the decomposed list of matrices is correct after the decomposition, I think it should be in the tests and not here. If we provide something for the user, we should make sure in advance that it is correct by testing it ourselves.

Unless there are some round-off error or special reasons that we cannot tell ahead if the decomposition will succeed?

purva-thakre · 2021-06-24T15:14:13Z

src/qutip_qip/decompositions/general_decompositions.py

+    if input_check_bool == True:
+        input_shape = input_gate.shape
+        if input_shape[0]%2==0.0: # won't be 2 for higher qudits
+        # TO DO : if d=4 (qu4it) then it will still appear as a d=2 i.e. qubit gate
+        # change to powers of 2
+            number_of_qubits = np.log(input_shape[0])/np.log(2)


@hodgestar @BoxiLi Is there a better way to find the number of qubits based on the size/shape of the input gate matrix ? Here, with what I have chosen to do, if the qudits are of dimensions that are power of 2 then my condition is valid for both qubit and said qudit.

Could dims work here ?

If I have a 2 qubit state unitary then dims = [[[2], [2]], [[2], [2]]] but if it was a qu4it state unitary then dims = [[4,4]].

And I think, a qutrit-qudit gate will also be incorrectly identified as a qubit gate. But I can change the line to be division by powers of 2 which will help catch odd-even dimensional states.

@BoxiLi @hodgestar I edited my previous comment. I think I have a better idea now. It's possible my understanding of dims is wrong.

Is there a reason that we need to consider qudits/qutrits? I thought the whole decomposition module assumes that we are working with qubits? Or is there anything also useful for qudits?

This kind of information is much easier to extract from the Qobj, before taking the full numpy matrix. A 2-qubit ket state has dims=[[2, 2], [1, 1]], while a qudit ket dims=[[4], [1]].

For unitaries acting on them: dims=[[2, 2], [2, 2]] and dims=[[4], [4]]

Is there a reason that we need to consider qudits/qutrits?

Not really. As the user can input any n x n unitary, I added this extra check to make sure the unitary acts only on qubits and not on other dimensions.

Plus, this should be able to find the number of qubits a unitary is acting on by looking at the shape/dims.

Ok, then this is just a validity check. Since the input is Qobj, using dims is easier. We can check if it is a list of 2.

input_gate.dims[0] == [2] * len(input_gate.dims[0])

Yeah, that's my problem. dims cannot distinguish between a 4 x 4 input. It could be an object acting on 2 qubits or a qudit of d = 4.

q = Qobj([[0., 0, 0, 0.],[0., 0, 0, 0.],[0., 0, 0, 0.],[0., 0, 0, 0.]]) q.dims = [[4], [4]]

I think I am going to change the input to the function. In addition to the input Qobj, they will also need to input number of qubits. This could help avoid the ambiguity. Then, this function could check if the shape of input gate is compatible with number of qubits.

Well, technically speaking, the example you provided is interpreted in QuTiP as a unitary acting on a qudit system, not two qubits.
E.g.

q = Qobj([[0., 0, 0, 0.],[0., 0, 0, 0.],[0., 0, 0, 0.],[0., 0, 0, 0.]], dims = [[4], [4]]) q * basis([2,2],[0,0])

will raise an error because the dimension does not match. This is how QuTiP distinguish them. If the user is meant to provide an operator acting on qubits, the dimension should match.

That said, in decomposition we can also be lenient and allow the user to just produce a 2^n times 2^n Qobj.

So I see two options:

We make it strict and all the input must have the dimension [[2,2,2,...],[2,2,2,...]

We allow it to be ambiguous a bit and require that input.shape[0] is a power of 2, by some smart bit manipulation: https://stackoverflow.com/questions/57025836/how-to-check-if-a-given-number-is-a-power-of-two. Although it will be a bit hard to find out what is n since technically log2(2^n) is not always reliable to give as n. Asking the user to provide the number of qubits is also fine.

purva-thakre · 2021-06-26T03:32:29Z

Closing this PR because I want to start on a new branch. When trying to rebase this branch, there were a lot of conflicts.

purva-thakre added 3 commits June 8, 2021 09:59

input type

fb01b4d

add files and folders

49a3fbf

normalize matrix

d48e75d

purva-thakre changed the title ~~Adds sungle qubit decomposition functions~~ Adds single qubit decomposition functions Jun 11, 2021

purva-thakre added 2 commits June 11, 2021 15:00

update tests

e4aa0f2

add check unitary

72055f9

purva-thakre commented Jun 11, 2021

View reviewed changes

tests/test_general_decompositions.py Outdated Show resolved Hide resolved

purva-thakre added 6 commits June 11, 2021 16:17

zero determinant

9ea0504

global phase start

d7adb86

change to 1d array test

2e0d10a

global phase part 1

0a7f660

real determinant

a7efe07

extract phase normalized array

bf4daed

purva-thakre commented Jun 13, 2021

View reviewed changes

src/qutip_qip/decompositions/general_decompositions.py Outdated Show resolved Hide resolved

purva-thakre commented Jun 13, 2021

View reviewed changes

src/qutip_qip/decompositions/general_decompositions.py Outdated Show resolved Hide resolved

purva-thakre commented Jun 13, 2021

View reviewed changes

BoxiLi reviewed Jun 14, 2021

View reviewed changes

purva-thakre added 8 commits June 14, 2021 11:17

remove functions accepting np array as input

6506bf9

extract phase

bc984d2

stash for install

d9d625a

input checks

c0e6ea2

stash for install

54ece0b

abc_zy

c35014f

s

b461727

gate imports

a4975af

BoxiLi added this to the GSoC 2021 Quantum Gate Decomposition milestone Jun 21, 2021

purva-thakre added 2 commits June 24, 2021 09:53

change gate import

f15d6f6

changes for ques

d11a425

purva-thakre commented Jun 24, 2021

View reviewed changes

purva-thakre closed this Jun 26, 2021

purva-thakre deleted the phase_su_decomposition branch June 26, 2021 03:39

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adds single qubit decomposition functions #52

Adds single qubit decomposition functions #52

purva-thakre commented Jun 11, 2021 •

edited

Loading

purva-thakre Jun 13, 2021 •

edited

Loading

BoxiLi Jun 14, 2021

purva-thakre Jun 13, 2021 •

edited

Loading

BoxiLi Jun 14, 2021

BoxiLi Jun 14, 2021

BoxiLi Jun 14, 2021

purva-thakre Jun 24, 2021 •

edited

Loading

purva-thakre Jun 24, 2021

BoxiLi Jun 24, 2021

BoxiLi Jun 24, 2021

purva-thakre Jun 24, 2021 •

edited

Loading

BoxiLi Jun 24, 2021

purva-thakre Jun 24, 2021 •

edited

Loading

BoxiLi Jun 24, 2021

BoxiLi Jun 24, 2021 •

edited

Loading

purva-thakre commented Jun 26, 2021


		import numpy as np

		def normalize_matrix(input_array)-> np.array:

Adds single qubit decomposition functions #52

Adds single qubit decomposition functions #52

Conversation

purva-thakre commented Jun 11, 2021 • edited Loading

purva-thakre Jun 13, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

purva-thakre Jun 13, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

purva-thakre Jun 24, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

purva-thakre Jun 24, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

purva-thakre Jun 24, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

BoxiLi Jun 24, 2021 • edited Loading

Choose a reason for hiding this comment

purva-thakre commented Jun 26, 2021

purva-thakre commented Jun 11, 2021 •

edited

Loading

purva-thakre Jun 13, 2021 •

edited

Loading

purva-thakre Jun 13, 2021 •

edited

Loading

purva-thakre Jun 24, 2021 •

edited

Loading

purva-thakre Jun 24, 2021 •

edited

Loading

purva-thakre Jun 24, 2021 •

edited

Loading

BoxiLi Jun 24, 2021 •

edited

Loading