This is an automated email from the ASF dual-hosted git repository.
git-site-role pushed a commit to branch asf-site
in repository https://gitbox.apache.org/repos/asf/datasketches-website.git
The following commit(s) were added to refs/heads/asf-site by this push:
new d854eb67 Automatic Site Publish by Buildbot
d854eb67 is described below
commit d854eb67f042f834d824bd7072810f5d51c2dfc2
Author: buildbot <[email protected]>
AuthorDate: Sun Jan 25 01:27:01 2026 +0000
Automatic Site Publish by Buildbot
---
output/docs/Density/DensitySketch.html | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/output/docs/Density/DensitySketch.html
b/output/docs/Density/DensitySketch.html
index dede78c7..7381535c 100644
--- a/output/docs/Density/DensitySketch.html
+++ b/output/docs/Density/DensitySketch.html
@@ -381,7 +381,7 @@ https://proceedings.mlr.press/v99/karnin19a/karnin19a.pdf
<h4 id="key-highlights">Key Highlights:</h4>
<ul>
<li><strong>New Complexity Measure:</strong> The authors define “class
discrepancy” as a way to characterize the coreset complexity of different
function families, similar to how Rademacher complexity is used for
generalization.</li>
- <li><strong>Improved Coreset Sizes:</strong> They prove the existence of
$\epsilon$-approximation coresets of size $O(\sqrt{d}/\epsilon)$ for several
common machine learning problems, including:
+ <li><strong>Improved Coreset Sizes:</strong> They prove the existence of
ε-approximation coresets of size <em>O(√d/ε)</em> for several common machine
learning problems, including:
<ul>
<li>Logistic regression</li>
<li>Sigmoid activation loss</li>
@@ -389,7 +389,7 @@ https://proceedings.mlr.press/v99/karnin19a/karnin19a.pdf
<li>Kernel density estimation</li>
</ul>
</li>
- <li><strong>Gaussian Kernel Resolution:</strong> The paper resolves a
long-standing open problem by matching the lower bound for the coreset
complexity of Gaussian kernel density estimation at $O(\sqrt{d}/\epsilon)$.</li>
+ <li><strong>Gaussian Kernel Resolution:</strong> The paper resolves a
long-standing open problem by matching the lower bound for the coreset
complexity of Gaussian kernel density estimation at <em>O(√d/ε)</em>.</li>
<li><strong>Streaming Algorithms:</strong> It introduces an exponential
improvement to the “merge-and-reduce” trick, leading to better streaming
sketches for any problem with low discrepancy.</li>
<li><strong>Deterministic Algorithm:</strong> The authors provide a simple,
deterministic algorithm for finding low-discrepancy sequences and coresets for
any positive semi-definite kernel.</li>
</ul>
@@ -399,7 +399,7 @@ https://proceedings.mlr.press/v99/karnin19a/karnin19a.pdf
<p>The findings allow for significantly faster optimization in large-scale
machine learning. By reducing a massive dataset into a much smaller coreset,
researchers can perform complex calculations (like training a logistic
regression model) with a fraction of the computational cost while maintaining a
high level of accuracy.</p>
<p><a id="inspiration"></a></p>
-<h3
id="our-implementations-was-inspired-by-the-following-implementation-example-and-tests-by-edo-liberty">Our
implementations was inspired by the following implementation, example, and
tests by Edo Liberty:</h3>
+<h3
id="our-implementations-was-inspired-by-the-following-code-example-and-tests-by-edo-liberty">Our
implementations was inspired by the following code, example, and tests by Edo
Liberty:</h3>
<ul>
<li><strong>Code:</strong>
https://github.com/edoliberty/streaming-quantiles/blob/f688c8161a25582457b0a09deb4630a81406293b/gde.py</li>
<li><strong>Example</strong>
https://github.com/edoliberty/streaming-quantiles/blob/f688c8161a25582457b0a09deb4630a81406293b/gde_example_usage.ipynb</li>
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
