From 1e9c929929598159c8c6d25e7cd51b2bef135c3e Mon Sep 17 00:00:00 2001 From: Axel Donath Date: Fri, 20 Dec 2024 17:47:53 -0500 Subject: [PATCH] Update README.md benchmark numbers --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index cedece4..f59e114 100644 --- a/README.md +++ b/README.md @@ -69,7 +69,7 @@ Here are some results from the benchmarks in the [examples/benchmarks](https://g | ----------------------------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------- | | ![Time vs. Number of Components](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-components-predict.png) | ![Time vs. Number of Samples](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-samples-predict.png) | ![Time vs. Number of Features](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-features-predict.png) | -For prediction the speedup is around 2-3x for varying number of components and features. For the number of samples the cross-over point is around O(10^4) samples. +For prediction the speedup is around 5-6x for varying number of components and features and ~50x speedup on the GPU. For the number of samples the cross-over point is around O(10^3) samples. ### Training Time @@ -77,4 +77,4 @@ For prediction the speedup is around 2-3x for varying number of components and f | ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- | | ![Time vs. Number of Components](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-components-fit.png) | ![Time vs. Number of Samples](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-samples-fit.png) | ![Time vs. Number of Features](https://raw.githubusercontent.com/adonath/gmmx/main/docs/_static/time-vs-n-features-fit.png) | -For training the speedup is around >10x on the same architecture and close to 100x speedup on the GPU. However there is no guarantee that it will converge to the same solution as Scikit-Learn. But there are some tests in the `tests` folder that compare the results of the two implementations which shows good agreement. +For training the speedup is around ~5-6x on the same architecture and ~50x speedup on the GPU. However there is no guarantee that it will converge to the same solution as Scikit-Learn. But there are some tests in the `tests` folder that compare the results of the two implementations which shows good agreement.