lanl-ansi/dwig

D-Wave Instance Generator (D-WIG)

dwig's Issues

Add bqp2gms converter

Doc Strings

Add proper doc strings to code.

Batch generation

When generating many cases accessing the QPU properties and setting up the data structures is a significant overhead. This should be broken into two steps, so batches of cases can be generated much faster.

Consider adding scaling term

So that parameters can be within coupler ranges but you can still get your objective value in the correct units.

Testing if a value is 0.0

Checker for values being equal to 0.0 should use an isclose() function.

Explore use of couplers via namedtuple

RAN Problem with Planted Ground State?

Will this yield interesting test cases?

• Pick spins at random
• select couplers and fields to be consistent with those spins

Optimal Annealing Time Test Cases

Implement the instance generation procedure from, https://arxiv.org/abs/1705.07452

No Generator Error Message

When the command dwig.py is run without a generator the error message should be more informative.

Consistent Gauge Transformation Flag

Standardize the parameters for random-gauge-transformation and simple-ground-state.

Implement Max Clique Problem Generator

dev is in the max-clq branch.

Move assertions to DWIGExceptions

Qbsolv Translator

Add a translator from bqp-json to the qbsolv data format.

Work around random.choices

This function is not supported in python 3.5, we should work around it so that the code is more portable.

-cs, -ss option can cause JSON creation to fail

When using the following command:

./dwig.py -pp -cs 0,4 0,5 1,4 1,5 const -cp 0.2 -f -0.5

dwig will crash, stating "a spin pair must be x,y integers". The parser is attempting to take an arbitrary number of qubit pairs, and is unable to differentiate the "const" subparser. This can easily be worked around by swapping the order of the -pp and -cs arguments, as shown:

./dwig.py -cs 0,4 0,5 1,4 1,5 -pp const -cp 0.2 -f -0.5

The cause of this issue is in the argument parsing, and there is not an immediately obvious fix that would allow passing arbitrary coupler or spin sets.

Test Case Names

Consider removing i from cases names as all outputs are now ising.

Make bqp-json a separate python package?

The primary function would be to store the schema file and have a dedicated validate method, which checks properties that are outside the scope of json schema.

This would also be a good place to put the bqp utility scripts, such as the format translators.

Improve Logging

Use real logs instead of just dumping everything to stderr.

Double check that spin to boolean transform is correct

It is clear that the transform is idempotent, at least.

Add bqp2mzn converter

Coefficient Truncation

Consider adding an option to Truncate floating point values based on the precision limitations of the QPU.

Add tests for bqp2qh.py

Runtime Slowness

This issue is discussed in the README. This will be resolved when SAPI is compatible with python 3.

GraphViz Support

Build a tool for converting bqp-json instances to pdfs using http://www.graphviz.org/

Mathematical Documentation

Find a good way to integrate the mathematical form of a B-QP into the BQP-JSON.md file.

FL Ground State

Using the ground state of all -1/0 can result in an unfair advantage for tree search solvers. The reason being that always branching to the low value first results in immediately finding OPT.

Translators and Reverse Edges

Make sure translators are robust to reverse edges.