From c1072f744a4471b5bee8fd945d007327d183a6b2 Mon Sep 17 00:00:00 2001
From: KulikDM <rogerplanck8@gmail.com>
Date: Sun, 18 Aug 2024 15:51:30 +0200
Subject: [PATCH] Added faster KDE and CDF generation using upsampling

---
 .pre-commit-config.yaml               |  2 +-
 pythresh/thresholds/thresh_utility.py | 34 +++++++++++++++++++++++----
 2 files changed, 31 insertions(+), 5 deletions(-)

diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index 1302720..ae5f28b 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -42,7 +42,7 @@ repos:
         name: Sort imports
 
   - repo: https://github.com/charliermarsh/ruff-pre-commit
-    rev: v0.5.6
+    rev: v0.6.1
     hooks:
       - id: ruff
         args: [--exit-non-zero-on-fix, --fix, --line-length=180]
diff --git a/pythresh/thresholds/thresh_utility.py b/pythresh/thresholds/thresh_utility.py
index cda6651..1e1ca5d 100644
--- a/pythresh/thresholds/thresh_utility.py
+++ b/pythresh/thresholds/thresh_utility.py
@@ -1,5 +1,6 @@
 import numpy as np
 import scipy.stats as stats
+from scipy.interpolate import interp1d
 from scipy.special import ndtr
 from sklearn.decomposition import TruncatedSVD
 from sklearn.utils import check_array
@@ -20,24 +21,49 @@ def cut(decision, limit):
     return labels
 
 
+def gen_interp(x, y):
+
+    interpolator = interp1d(x, y, kind='cubic',
+                            fill_value='extrapolate')
+
+    return interpolator
+
+
 def gen_kde(data, lower, upper, size):
 
+    insize = min(size, 5000)
+
     # Create a KDE of the data
     kde = stats.gaussian_kde(data)
-    dat_range = np.linspace(lower, upper, size)
+    dat_range = np.linspace(lower, upper, insize)
+    dat_eval = np.linspace(lower, upper, size)
 
-    return kde(dat_range), dat_range
+    # Use interpolation for fast KDE upsampling
+    if size > insize:
+        interpolator = gen_interp(dat_range, kde(dat_range))
+        return interpolator(dat_eval), dat_eval
+
+    return kde(dat_eval), dat_eval
 
 
 def gen_cdf(data, lower, upper, size):
 
+    insize = min(size, 5000)
+
     # Create a KDE & CDF of the data
     kde = stats.gaussian_kde(data)
-    dat_range = np.linspace(lower, upper, size)
+    dat_range = np.linspace(lower, upper, insize)
+    dat_eval = np.linspace(lower, upper, size)
+
     cdf = np.array(tuple(ndtr(np.ravel(item - kde.dataset) / kde.factor).mean()
                          for item in dat_range))
 
-    return cdf, dat_range
+    # Use interpolation for fast CDF upsampling
+    if size > insize:
+        interpolator = gen_interp(dat_range, cdf)
+        return interpolator(dat_eval), dat_eval
+
+    return cdf, dat_eval
 
 
 def check_scores(decision, random_state=1234):