Ctrl+K

Penalized logP (using lambo)

Penalized logP (using lambo)#

Type of objective function: discrete Environment to run this objective function: poli lambo

About#

This objective function computes the penalized logP exactly as is done in the lambo implementation1 [Stanton et al., 2022].

To do so, we import their scoring function.

Prerequisites#

None. This black box should run out-of-the-box.

How to run#

import numpy as np
from poli.objective_repository import (
    PenalizedLogPLamboProblemFactory,
    PenalizedLogPLamboBlackBox,
)

# Creating the black box
f = PenalizedLogPLamboBlackBox()

# Creating a problem
problem = PenalizedLogPLamboProblemFactory().create()
f, x0 = problem.black_box, problem.x0

# Example input: a single carbon
x = np.array(["C"]).reshape(1, -1)

# Querying:
y = f(x)
print(y)  # Should be close to -6.2238

How to cite#

If you use this black box, we expect you to cite the following resources:

[1] Stanton, Samuel, Wesley Maddox, Nate Gruver, Phillip Maffettone, Emily Delaney, Peyton Greenside, and Andrew Gordon Wilson. “Accelerating Bayesian Optimization for Biological Sequence Design with Denoising Autoencoders.” arXiv, July 12, 2022. http://arxiv.org/abs/2203.12742.

[2] González-Duque, M., Bartels, S., & Michael, R. (2024). poli: a libary of discrete sequence objectives [Computer software]. MachineLearningLifeScience/poli


@article{stanton:LaMBO:2022,
  title   = {Accelerating Bayesian Optimization for Biological Sequence Design with Denoising Autoencoders},
  author  = {Stanton, Samuel and Maddox, Wesley and Gruver, Nate and Maffettone, Phillip and Delaney, Emily and Greenside, Peyton and Wilson, Andrew Gordon},
  journal = {arXiv preprint arXiv:2203.12742},
  year    = {2022}
}

@software{Gonzalez-Duque:poli:2024,
author = {González-Duque, Miguel and Bartels, Simon and Michael, Richard},
month = jan,
title = {{poli: a libary of discrete sequence objectives}},
url = {https://github.com/MachineLearningLifeScience/poli},
version = {0.0.1},
year = {2024}
}
1

If you check carefully, you might have noticed that they add to their implementation the empirical means and standard deviations of the ZINC dataset for the values they compute.

Contents