Jakob Jordan

Always use `reorder=True` and remove possiblity to configure this?

2

Is there a downside to reordering the genome? If not I think we should remove the possibility of choosing this setting and just always doing it. We need to keep...

At the moment one is forced to implement all `to_...` functions for a new node. However, in some cases one may be happy with just one, e.g., `to_func` bc one...

At the moment one can not print an individual before its fitness (for any objective) is determined. Ultimately this is caused by, e.g., this call sequence `IndividualSingleGenome::__repr__->IndividualBase::fitness` in which the...

Assert runtime invariant: all individuals in `MuPlusLambda::step` need to have a non-None index

3

The logic in `MuPlusLambda` (and likely in other places) assumes that all individuals in a population have been assigned a valid index. We should thus make sure of this in...

(as the title suggests) closes #375

Currently, it's only possible to define hurdles like "only the best 50% of the population passes". However, it may also be useful to be able to define hurdles like "only...

At the moment we define specific versions for all dependencies. This can lead to issues (see, e.g., https://github.com/Happy-Algorithms-League/hal-cgp/pull/368/) besides potentially causing issues for users. We should unfreeze the dependencies and...

Using a MSE error function can cause GP to prefer solutions which have wrong functional form but, for example, are close to the mean of the target data over solutions...

Currently, we merely make sure that the examples don't crash, not that they lead to the correct results. We should add strategic `assert`s, warning us if an example does not...

As indicated by recent events :eyes: , we should include MacOS in our CI pipeline to make sure everything works as expected also on such systems.