Script to run the HEBO algorithm on a BBOB function.

The change from huawei-noah/HEBO#61 (comment) is integrated here.

HEBO/.gitignore

@@ -19,3 +19,5 @@ worksapce/
 
 # catboost
 catboost_info/
+
+.conda_environment

HEBO/INSTALL

@@ -0,0 +1,8 @@
+#!/usr/bin/env fish
+
+true
+and conda activate base
+and rm -rf ./.conda_environment
+and conda env create --prefix ./.conda_environment --file conda.yaml
+and conda activate ./.conda_environment
+and pip install --default-timeout=300 -r requirements.txt

HEBO/conda.yaml

@@ -0,0 +1,5 @@
+channels:
+  - defaults
+dependencies:
+  - pip=23.3
+  - python=3.9.18

HEBO/hebo/models/gp/gp_util.py

@@ -10,14 +10,14 @@
 import numpy as np
 import torch
 import torch.nn as nn
-from gpytorch.kernels import (AdditiveKernel, MaternKernel, ProductKernel,
-                              ScaleKernel)
-from gpytorch.priors import GammaPrior
 from torch import FloatTensor, LongTensor
 
-from ..layers import EmbTransform
-from ..util import get_random_graph
 
+from gpytorch.kernels import MaternKernel, ScaleKernel, ProductKernel
+from gpytorch.priors  import GammaPrior
+from gpytorch.constraints.constraints import LessThan
+
+from ..layers import EmbTransform
 
 class DummyFeatureExtractor(nn.Module):
     def __init__(self, num_cont, num_enum, num_uniqs = None, emb_sizes = None):
@@ -43,7 +43,8 @@ def default_kern(x, xe, y, total_dim = None, ard_kernel = True, fe = None, max_x
         kerns    = []
         if has_num:
             ard_num_dims = x.shape[1] if ard_kernel else None
-            kernel       = MaternKernel(nu = 1.5, ard_num_dims = ard_num_dims, active_dims = torch.arange(x.shape[1]))
+            kernel       = MaternKernel(nu = 1.5, ard_num_dims = ard_num_dims, active_dims = torch.arange(x.shape[1]),
+                                        lengthscale_constraint=LessThan(5))
             if ard_kernel:
                 lscales = kernel.lengthscale.detach().clone().view(1, -1)
                 for i in range(x.shape[1]):
@@ -52,19 +53,21 @@ def default_kern(x, xe, y, total_dim = None, ard_kernel = True, fe = None, max_x
                 kernel.lengthscale = lscales
             kerns.append(kernel)
         if has_enum:
-            kernel = MaternKernel(nu = 1.5, active_dims = torch.arange(x.shape[1], total_dim))
+            kernel = MaternKernel(nu = 1.5, active_dims = torch.arange(x.shape[1], total_dim),
+                                        lengthscale_constraint=LessThan(5))
             kerns.append(kernel)
         final_kern = ScaleKernel(ProductKernel(*kerns), outputscale_prior = GammaPrior(0.5, 0.5))
         final_kern.outputscale = y[torch.isfinite(y)].var()
         return final_kern
     else:
         if ard_kernel:
-            kernel = ScaleKernel(MaternKernel(nu = 1.5, ard_num_dims = total_dim))
+            kernel = ScaleKernel(MaternKernel(nu = 1.5, ard_num_dims = total_dim,
+                                        lengthscale_constraint=LessThan(5)))
         else:
             kernel = ScaleKernel(MaternKernel(nu = 1.5))
         kernel.outputscale = y[torch.isfinite(y)].var()
         return kernel
-    
+
 def default_kern_rd(x, xe, y, total_dim = None, ard_kernel = True, fe = None, max_x = 1000, E=0.2):
     '''
     Get a default kernel with random decompositons. 0 <= E <=1 specifies random tree conectivity.

HEBO/hebo_on_bbob.py

@@ -0,0 +1,71 @@
+# -*- coding: utf-8 -*-
+"""HEBO_on_BBOB.ipynb
+
+Automatically generated by Colaboratory.
+
+Original file is located at
+    https://colab.research.google.com/drive/1XftMKU7-tWj0cdWjH7XsfiDPBIKXAWZk
+"""
+
+import hebo
+import ioh
+import numpy
+
+ioh_logger = ioh.logger.Analyzer(
+  [ioh.logger.trigger.ALWAYS],
+  additional_properties = [
+    # ioh.logger.property.EVALUATIONS,
+    # ioh.logger.property.RAWY,
+    ioh.logger.property.RAWYBEST,
+    ioh.logger.property.TRANSFORMEDY,
+    ioh.logger.property.TRANSFORMEDYBEST,
+    # ioh.logger.property.CURRENTBESTY,
+    # ioh.logger.property.CURRENTY,
+    # ioh.logger.property.PENALTY,
+    # ioh.logger.property.VIOLATION,
+  ],
+  algorithm_name = "HEBO",
+  folder_name = "HEBO",
+  root = "HEBO_on_BBOB",
+)
+
+# WARNING: The initial states of the tool to generate random numbers are NOT fixed within these parameters.
+BBOB_SEARCH_SPACE_LOWER_BOUND = -5
+BBOB_SEARCH_SPACE_UPPER_BOUND = 5
+num_variables_list = [60]
+fids = [1]
+instances = [0]
+num_runs = 10
+run_budget = 80
+total_num_runs = num_runs * len(instances) * len(fids) * len(num_variables_list)
+run_index = 0
+for fid in fids:
+  for instance in instances:
+    for num_variables in num_variables_list:
+      for run_index in range(num_runs):
+        bbob_problem = ioh.get_problem(
+          fid = fid,
+          instance = instance,
+          dimension = num_variables,
+        )
+        bbob_problem.attach_logger(ioh_logger)
+
+        hebo_space = hebo.design_space.design_space.DesignSpace().parse([
+          {
+            'name': f'x{variable_index}',
+            'type': 'num',
+            'lb': BBOB_SEARCH_SPACE_LOWER_BOUND,
+            'ub': BBOB_SEARCH_SPACE_UPPER_BOUND,
+          }
+          for variable_index in range(num_variables)
+        ])
+
+        hebo_optimizer = hebo.optimizers.hebo.HEBO(hebo_space)
+        for i in range(run_budget):
+          recommendations_DataFrame = hebo_optimizer.suggest(n_suggestions = 1)
+          recommendations_list = recommendations_DataFrame.iloc[0].tolist()
+          recommendation = numpy.array([bbob_problem(recommendations_list)])
+          hebo_optimizer.observe(recommendations_DataFrame, recommendation)
+
+        run_index += 1
+        print(f"Done: {run_index}/{total_num_runs:,}")

HEBO/requirements.txt

@@ -1,11 +1,8 @@
-numpy>=1.16,<1.25
-pandas>=1.0.1
-torch>=1.9.0
-pymoo>=0.6.0
-scikit-learn>=0.22
-gpytorch>=1.4.0
-GPy>=1.9.9
-catboost>=0.24.4
-xgboost
-lightgbm
-disjoint-set
+disjoint-set==0.7.4
+gpytorch==1.11
+ioh==0.3.12
+jupyter==1.0.0
+pandas==2.1.3
+pymoo==0.6.0.1
+scikit-learn==1.3.2
+torch==2.1.1