QubitizationWalkOperator add attribute: sum of LCU coefficients#890
Conversation
|
|
||
| select: SelectOracle | ||
| prepare: PrepareOracle | ||
| sum_of_ham_coeffs: Optional[float] |
There was a problem hiding this comment.
The default value here should be 1 or something else?
There was a problem hiding this comment.
I didn't set a default to avoid accidentally forgetting it while building walk ops
But as the sum_of_ham_coeffs attribute is not needed in the decomposition/call-graph, it won't break anything. It's only needed for Hamiltonian simulation, in which case it needs to be correctly calculated anyway.
There was a problem hiding this comment.
right, I think Optional[float] = None in that case.
There was a problem hiding this comment.
would it be better to not have a default, and let the user explicitly specify None when they don't know the value? (just to make sure it's not accidentally omitted)
|
It's typically called lambda in the literature which is problematic in python as a variable name I know. Maybe sum_of_lcu_coeffs? |
|
Is |
anurudhp
force-pushed
the
2024-04-qwalkop-scale-factor
branch
from
c8c6269 to
aff2de8
Compare
|
Yes I think that makes sense. That way we can also implicitly compute this value because we usually pass the exact coeffs to Prepare (which are normalized there). For ex. it would simplify this (we can drop L93) Qualtran/qualtran/bloqs/chemistry/ising/hamiltonian.py Lines 92 to 94 in 91b80b1 |
anurudhp
force-pushed
the
2024-04-qwalkop-scale-factor
branch
from
cc39c78 to
cde48eb
Compare
|
@tanujkhattar PTAL |
anurudhp
force-pushed
the
2024-04-qwalkop-scale-factor
branch
from
bd58405 to
d84fde2
Compare
anurudhp
force-pushed
the
2024-04-qwalkop-scale-factor
branch
from
cf873a9 to
1d1612f
Compare
|
|
||
| @cached_property | ||
| def sum_of_lcu_coefficients(self) -> Optional[float]: | ||
| r"""value of $\lambda$, i.e. sum of the coefficients $w_l$.""" |
There was a problem hiding this comment.
Let's clarify by writing "sum of absolute values of coefficients. " instead. In this case, the docstring of the class does clarify that each
| a selection register $|i\rangle$. In order to prepare such a state, the PREPARE circuit is also | ||
| allowed to use a junk register that is entangled with selection register. | ||
| the coefficients as amplitudes of a state $|\Psi\rangle = \sum_{i=0}^{N-1} \sqrt{\frac{c_{i}}{\lambda}} |i\rangle$ | ||
| where $\lambda = \sum_i c_i$, using a selection register $|i\rangle$. In order to prepare such |
There was a problem hiding this comment.
| where $\lambda = \sum_i c_i$, using a selection register $|i\rangle$. In order to prepare such | |
| where $\lambda = \sum_i | c_i |$, using a selection register $|i\rangle$. In order to prepare such |
|
|
||
| @property | ||
| def l1_norm_of_coeffs(self) -> Optional['SymbolicFloat']: | ||
| r"""Sum of the coefficients $c_i$. |
There was a problem hiding this comment.
| r"""Sum of the coefficients $c_i$. | |
| r"""Sum of the absolute values of coefficients $c_i$. |
3 participants
fixes #868
The sum of coefficients determines the complexity of hamiltonian simulation. This will also enable us to propagate symbolic values to the prepare gate.
@tanujkhattar could you review this? Also it would be great if someone with qchem expertise checks whether the sum of coeffs for the Hubbard gate is correct.