Index Kernel

[AdrianSosic]

DevPR, parent is #745

Adds the IndexKernel and PositiveIndexKernel classes

The new factory design added in #746 offers full customization of the GP, following a composition approach. To reach the full potential that this offers, having BayBE IndexKernel classes to let the user to construct custom architectures is the natural next step, in particular because there other reasonable choices than just having the standard ICM product, such as:

• Selecting the type of index kernel (e.g. regular vs positive)
• Implementing multi-fidelity approaches (see Add DiscreteNumericalFidelityParameter and CategoricalFidelityParameter #715)
• Creating kernels that can handle multiple task parameters

In addition:

• It finally provides an interface for configuring the kernel itself, e.g. its rank (which used to be a TODO for ages)
• And it offers the new parameter-named BayBE for conveniently specifying active dimensions for the above tasks.

[Copilot]

Pull request overview

Adds an IndexKernel and expands GaussianProcessSurrogate customization via kernel/mean/likelihood factories, including new GP presets and serialization behavior for non-BayBE GP components.

Changes:

• Introduces IndexKernel and adds it to Hypothesis kernel strategies.
• Refactors GP surrogate to accept kernel/mean/likelihood components (including some GPyTorch objects), adds EDBO / EDBO_SMOOTHED presets, and updates preset exports.
• Updates serialization hooks and generic/abstract type handling; adds tests for transfer learning and GPyTorch-kernel behavior.

Reviewed changes

Copilot reviewed 23 out of 23 changed files in this pull request and generated 12 comments.

Show a summary per file

File	Description
tests/test_transfer_learning.py	Adds transfer-learning recommendation test coverage
tests/test_gp.py	Adds tests for using GPyTorch kernels and blocking their serialization
tests/hypothesis_strategies/kernels.py	Extends kernel generation strategies to include IndexKernel
baybe/utils/boolean.py	Updates abstract detection to handle generics
baybe/utils/basic.py	Updates subclass discovery to handle generics
baybe/surrogates/gaussian_process/presets/utils.py	Adds lazy-loaded constant mean factory
baybe/surrogates/gaussian_process/presets/edbo_smoothed.py	Adds “smoothed EDBO” kernel/likelihood preset implementations
baybe/surrogates/gaussian_process/presets/edbo.py	Refactors EDBO preset into kernel/mean/likelihood factories
baybe/surrogates/gaussian_process/presets/default.py	Repoints default preset to smoothed EDBO kernel/likelihood + lazy mean
baybe/surrogates/gaussian_process/presets/core.py	Adds preset enum values and preset-based GP construction
baybe/surrogates/gaussian_process/presets/init.py	Expands exported preset factories and core API
baybe/surrogates/gaussian_process/kernel_factory.py	Removes old kernel factory module (replaced by components)
baybe/surrogates/gaussian_process/core.py	Adds mean/likelihood factories; uses component factories and IndexKernel for multitask
baybe/surrogates/gaussian_process/components/mean.py	Introduces GP mean component factory typing
baybe/surrogates/gaussian_process/components/likelihood.py	Introduces GP likelihood component factory typing
baybe/surrogates/gaussian_process/components/kernel.py	Introduces GP kernel component factory typing
baybe/surrogates/gaussian_process/components/generic.py	Adds generic component factory + serialization blocking hook for GPyTorch kernels
baybe/surrogates/gaussian_process/components/init.py	Exposes component factory types
baybe/serialization/core.py	Improves base-class (un)structuring and non-BayBE type names in error messages
baybe/searchspace/core.py	Refactors task-parameter lookup into a helper property
baybe/kernels/basic.py	Adds IndexKernel
baybe/kernels/base.py	Updates kernel factory import path and broadens active_dims type
CHANGELOG.md	Documents new GP customization support, presets, and IndexKernel

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[AVHopp]

LGTM, only one minor question regarding the design