I have a related question. Recently I have been experimenting on the Pegasus machine with a scalable, fully-connected problem. Using 30 logical qubits yields an embedding with maximum chain lengths of 5 and the anneal result is outstanding. However when I scaled it up to 75 logical qubits, the embedding has chain lengths ranging from 8-10 and the results were all over the place. The inspector gave warnings that these chain lengths are greater than 7 so I would assume it was part of the issue.
I'm using this function to generate embeddings for my graph. I wonder do you have any suggestion on how should I approach this type of problems?
Chains are often required in larger problems. Here is a post that elaborates on this in detail. The warning you see from the Inspector is a caution message.
In order to improve the results for your problem, you might want to start by analyzing the chain_break_fraction in your results. A chain break occurs when a chain of qubits representing a variable end up returning non-identical values.
If you notice there is high chain_break_fraction, you can increase the chain_strength. A good value to start with for chain_strength is the largest absolute bias value in your problem’s equation. Here is a paper that explains the importance of this parameter and an example that resolves the chain break issue in a problem.
Comments
Hello,
We have this paper about setting chain strength that might be helpful:
https://www.dwavesys.com/sites/default/files/14-1041A-A_Setting_The_Chain_Strength.pdf
Can you give any further context? What kind of information are you after?
We can try to provide more if needed!
Hopefully this was helpful.
Don't hesitate to ask further questions!
I have a related question. Recently I have been experimenting on the Pegasus machine with a scalable, fully-connected problem. Using 30 logical qubits yields an embedding with maximum chain lengths of 5 and the anneal result is outstanding. However when I scaled it up to 75 logical qubits, the embedding has chain lengths ranging from 8-10 and the results were all over the place. The inspector gave warnings that these chain lengths are greater than 7 so I would assume it was part of the issue.
I'm using this function to generate embeddings for my graph. I wonder do you have any suggestion on how should I approach this type of problems?
Hi Hanjing,
Chains are often required in larger problems. Here is a post that elaborates on this in detail. The warning you see from the Inspector is a caution message.
In order to improve the results for your problem, you might want to start by analyzing the chain_break_fraction in your results. A chain break occurs when a chain of qubits representing a variable end up returning non-identical values.
If you notice there is high chain_break_fraction, you can increase the chain_strength. A good value to start with for chain_strength is the largest absolute bias value in your problem’s equation. Here is a paper that explains the importance of this parameter and an example that resolves the chain break issue in a problem.
Please let me know if the issue persists.
Please sign in to leave a comment.