evojax VS cuml

Awesome results! Would love to see a comparison with other accelerated evolution methods (eg https://github.com/google/evojax)

For which problems do you want to apply multiple-objective optimization? What are the objectives? https://mml-book.github.io/ is only about single objective optimization (including constraints). Multi objective reinfocement learning (https://arxiv.org/pdf/1908.08342.pdf) is not really MO-optimization. Machine learning optimization frameworks like https://github.com/google/evojax support single objective optimization and quality-diversity (MAP-elites), but not MO-optimization.

population based algos dont care about differentiability https://github.com/google/evojax https://github.com/RobertTLange/evosax https://github.com/nnaisense/evotorch https://github.com/uber-research/PyTorch-NEAT

Googles evojax project shows that evolutionary algorithms may be applied in the machine learning domain. And https://github.com/google/jax provides means to implement these algorithms to be deployed on CPUs/GPUs or even TPUs. But some questions remain unanswered:

Check out EvoJAX if you haven't seen it! Recently released for neuroevolution

Code for https://arxiv.org/abs/2202.05008 found: https://github.com/google/evojax

sklearn can't, bit take a look at cuML (https://github.com/rapidsai/cuml ). It uses the same API as sklearn but executes on GPU.

As a companion to the other comments, I'd like to mention that the RAPIDS library cuML provides GPU-accelerated versions of quite a few of the algorithms mentioned in this thread (HDBSCAN, UMAP, SVM, PCA, {Exact, Approximate} Nearest Neighbors, DBSCAN, KMeans, etc.).

CuML

Did they implement the clustering algorithm themselves? cuML is a GPU-accelerated scikit-learn-like package that covers many of the common ML algorithms.

https://github.com/rapidsai/cuml

> cuML is a suite of libraries that implement machine learning algorithms and mathematical primitives functions that share compatible APIs with other RAPIDS projects. cuML enables data scientists, researchers, and software engineers to run traditional tabular ML tasks on GPUs without going into the details of CUDA programming. In most cases, cuML's Python API matches the API from scikit-learn. For large datasets, these GPU-based implementations can complete 10-50x faster than their CPU equivalents. For details on performance, see the cuML Benchmarks Notebook.

Cython code

Take a look at RAPIDS CUML https://github.com/rapidsai/cuml. It's useful for most common ML algorithms. Feel free to create Github issues for feature requests & bugs.