Starting part 2!

[TumCucTom]

We have a score of 0.73 for the second challenge by doing the following:

We hope that this will help others get started with the second part of the challenge.

Please let us know if this works for you!!

[bdavidzhang]

Is this for graph 4?

[TumCucTom]

Our score of 0.73 is for graph 4, yes

[TumCucTom]

Is this for graph 4?

let us know if this works for you

[vijeycreative]

Our score of 0.73 is for graph 4, yes

Hey, is this the balanced score? Can I ask what did you set your lambda value and your learning rate? I can't seem to cross 0.5.

[TumCucTom]

we have never modified learning rate

think i mentioned in another channel but our lambda here isn't sensitive - consistent results of 0.78-0.79 with lambda between 5-20

[TumCucTom]

with less than a minute of learning time for graph 4

[bdavidzhang]

Are you summing over the edges for $n$?

[TumCucTom]

not sure i understand the question?

[bdavidzhang]

So when you're implementing this solution, are the i and j indexes coming from the list of edges or the list of nodes?

[TumCucTom]

if you're talking about the i and j at the bottom that is the sum of Pauli-Z interactions between all pairs of qibits (each qibit represents a node)

so i think the answer to your question is nodes - but still slightly unsure if that's what you're asking about. Happy to try and provide further clarification

[SidCh1]

We have a score of 0.73 for the second challenge by doing the following:

We hope that this will help others get started with the second part of the challenge.

Please let us know if this works for you!!

We are trying to do the same but we added a parameter (lambda) so that we can trade of between max cut and subset sizes.

[SidCh1]

Also why power 2? Because Abs is not differentiable at 0?

[TumCucTom]

Also why power 2? Because Abs is not differentiable at 0?

exactly, that abs is hard to implement and that quantum var algos should benefit from 'nice' energies when differentiated

[TumCucTom]

We have a score of 0.73 for the second challenge by doing the following:

We hope that this will help others get started with the second part of the challenge.
Please let us know if this works for you!!

We are trying to do the same but we added a parameter (lambda) so that we can trade of between max cut and subset sizes.

yeah we are currently trying to use a lamda to get a reasonable solution for part 3 - don't know why with didn't back track and also use this to optimise part 2 further. Thanks!

[light-oishikk]

[TumCucTom]

Hi, not sure whether this is within the rules of the challenge - maybe confer with the IonQ guys first

[mulliganaceous]

Seriously, is it that easy? It took like a few hours for me to first expand in terms of X, and then convert it back to Z.