Bruno Kacper Mlodozeniec

Contrary to the interface of the other acquisitions, the `evaluate` and `evaluate_with_gradients` methods of `MultipointExpectedImprovement` acquisition return the _negated_ acquisition value and gradient. Example in `evaluate`: https://github.com/EmuKit/emukit/blob/9f6763f8a0b689b0cd05b64dbdbe27ba6d7953c3/emukit/bayesian_optimization/acquisitions/expected_improvement.py#L185 and in `evaluate_with_gradients`:...

Add gradient calculation for the covariance between points in GPyModelWrapper

12

This change adds a method `get_covariance_between_points_gradients` to the `GPyModelWrapper` to facilitate obtaining the gradients of the corresponding `get_covariance_between_points` method implemented in the wrapper. These gradients are useful when optimising any...

The proper way of combining HMC with batch acquisition strategies for Bayesian Optimization

5

This issue is meant to ask two questions: (1) Does the combination of Local Penalization and `IntegratedHyperparameterAcquisition` behave in the expected way, and (1) What's the right way to approach...

`fast-jl` doesn't work with `pip` version of 23 and above

2

If you try to install `fast-jl` or `traker[fast]` with `pip==23.3.2` or `pip==24.0`, you get the following error: ``` Collecting fast-jl==0.1.2 Using cached fast_jl-0.1.2.tar.gz (5.2 kB) Installing build dependencies ... done...

`vmap` with `scatter_add` extremely slow when using `xla_gpu_deterministic_ops`

4

### Description The issue is 1) about a rather significant slow-down to the `scatter_add` operation when running jax with the `xla_gpu_deterministic_ops=true` flag, and 2) about a further disproportionately large slow-down...

TRAK and fast-jl use deprecated `setup.py install` method

4

Installig `fast-jl` with Python 3.11 and `setuptools==65.5.0` gives: ``` > pip install fast-jl ... DEPRECATION: fast-jl is being installed using the legacy 'setup.py install' method, because it does not have...

Stop overriding `model_func` state dictionary when loading state dicts of K-FAC or K-FAC inverse

3

Currently, when calling `.load_state_dict(state_dict)` on a `KFACLinearOperator` or `KFACInverseLinearOperator` object initialised with some `model_func`, this will override the state dictionary of the `model_func`. This might lead to unexpected behaviour, wherein:...

The code for actually adding a shared equal contributions notice seems to only be present in the ICML template. In the NeurIPS template, setting `equal: true` for an author has...