Update: seeding

Author: heptaflar

bin/kaleidoscope-esa-cci-oc

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+#
+# Copyright (c) Brockmann Consult GmbH, 2025
+# License: MIT
+#
+set -e
+## Produces Monte Carlo variants of the ESA CCI Ocean Colour product.
+#
+#  ./kaleidoscope-esa-cci-oc <file>
+#
+for selector in 001 002 003 004 005 006 007 008 009 010; do
+  echo "$(tput setaf 2)$(date -u "+%Y-%m-%dT%H:%M:%S") [INFO] Selector ${selector} ...$(tput sgr0)"
+  kaleidoscope \
+    --selector ${selector} \
+    --product-type esa-cci-oc \
+    --log-level warning \
+    --progress \
+    "${1}" \
+    "${1%.nc}"."${selector}".nc
+done

bin/kaleidoscope-esa-scope-ex

@@ -8,12 +8,12 @@ set -e
 #
 #  ./kaleidoscope-esa-scope-ex
 #
-for selector in 003 005 007 011 013 017 019 023 029 031; do
+for selector in 001 002 003 004 005 006 007 008 009 010; do
   echo "$(tput setaf 2)$(date -u "+%Y-%m-%dT%H:%M:%S") [INFO] Selector ${selector} ...$(tput sgr0)"
   kaleidoscope \
     --selector ${selector} \
     --product-type esa-scope-exchange \
-    --log-level off \
+    --log-level warning \
     --progress \
     Ford_et_al_UExP-FNN-U_physics_carbonatesystem_ESASCOPE_v5.nc \
     Ford_et_al_UExP-FNN-U_physics_carbonatesystem_ESASCOPE_v5.${selector}.nc

bin/kaleidoscope-ghrsst

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+#
+# Copyright (c) Brockmann Consult GmbH, 2025
+# License: MIT
+#
+set -e
+## Produces Monte Carlo variants of the GHRSST product.
+#
+#  ./kaleidoscope-ghrsst <file>
+#
+for selector in 001 002 003 004 005 006 007 008 009 010; do
+  echo "$(tput setaf 2)$(date -u "+%Y-%m-%dT%H:%M:%S") [INFO] Selector ${selector} ...$(tput sgr0)"
+  kaleidoscope \
+    --selector ${selector} \
+    --product-type ghrsst \
+    --log-level warning \
+    --progress \
+    "${1}" \
+    "${1%.nc}"."${selector}".nc
+done

bin/kaleidoscope-glorys

@@ -0,0 +1,20 @@
+#!/usr/bin/env bash
+#
+# Copyright (c) Brockmann Consult GmbH, 2025
+# License: MIT
+#
+set -e
+## Produces Monte Carlo variants of the GLORYS product.
+#
+#  ./kaleidoscope-glorys <file>
+#
+for selector in 001 002 003 004 005 006 007 008 009 010; do
+  echo "$(tput setaf 2)$(date -u "+%Y-%m-%dT%H:%M:%S") [INFO] Selector ${selector} ...$(tput sgr0)"
+  kaleidoscope \
+    --selector ${selector} \
+    --product-type esa-cci-oc \
+    --log-level warning \
+    --progress \
+    "${1}" \
+    "${1%.nc}"."${selector}".nc
+done

kaleidoscope/algorithms/randomize.py

@@ -76,8 +76,8 @@ class Randomize(InformedBlockAlgorithm):
 
     def __init__(
         self,
-        dtype: np.dtype,
-        m: int,
+        dtype: np.dtype = np.single,
+        m: int = 2,
         dist: Literal["normal", "lognormal", "chlorophyll"] | str = "normal",
         entropy: int | list[int] | None = None,
     ):

kaleidoscope/operators/randomizeop.py

@@ -18,6 +18,7 @@
 from ..algorithms.codec import Decode
 from ..algorithms.codec import Encode
 from ..algorithms.randomize import Randomize
+from ..generators import DefaultGenerator
 from ..interface.logging import Logging
 from ..interface.operator import Operator
 from ..logger import get_logger
@@ -90,6 +91,10 @@ def run(self, source: Dataset) -> Dataset:  # noqa: D102
         :param source: The source dataset.
         :return: The result dataset.
         """
+        source_id = source.attrs.get(
+            "tracking_id",
+            source.attrs.get("uuid", self._args.source_file.stem),
+        )
         target = Dataset(
             data_vars=source.data_vars,
             coords=source.coords,
@@ -106,10 +111,9 @@ def run(self, source: Dataset) -> Dataset:  # noqa: D102
 
             a: dict[str:Any] = config[v]
             f = Randomize(
-                np.single,
-                x.ndim,
+                m=x.ndim,
                 dist=a["distribution"],
-                entropy=self.entropy(v),
+                entropy=self.entropy(v, source_id),
             )
             if "uncertainty" in a:
                 u = (
@@ -179,15 +183,20 @@ def config(self) -> dict[str : dict[str:Any]]:
                 config = json.load(r)
         return config
 
-    def entropy(self, v: str) -> list[int]:
+    def entropy(self, vid: str, did: str, n: int = 8) -> list[int]:
         """
         Returns the entropy of the seed sequence used for a given variable.
 
-        :param v: The name of the variable.
+        Entropy is generated using the Philox bit generator, which produces
+        truly independent sequences for different values of the seed.
+
+        :param vid: The variable ID.
+        :param did: The dataset ID.
+        :param n: The length of the seed sequence.
         :return: The entropy.
         """
-        return [
-            self._args.selector,
-            _hash(v),
-            _hash(self._args.source_file.stem),
-        ]
+        from numpy.random import Philox
+
+        seed = _hash(f"{vid}-{did}") + self._args.selector
+        g = DefaultGenerator(Philox(seed))
+        return [g.next() for _ in range(n)]