Release 1.0.2

CHANGELOG.md

@@ -2,6 +2,16 @@
 
 All notable changes to this project will be documented in this file.
 
+## [1.0.2] - 2026-04-08
+
+### Bug Fixes
+
+- Correct TI evaluation and leave out uncertainty reporting. Fixes [#81].
+
+### Miscellaneous Tasks
+
+- Change email addresses.
+
 ## [1.0.1] - 2025-09-10
 
 ### Bug Fixes
@@ -915,6 +925,7 @@ returns `None` instead. Fixes [#11]
 ## [0.5.3] - 2022-08-22
 
 <!-- markdownlint-disable-file MD024 -->
+[1.0.2]: https://github.com/lycosystem/lyscripts/compare/1.0.1...1.0.2
 [1.0.1]: https://github.com/lycosystem/lyscripts/compare/1.0.0...1.0.1
 [1.0.0]: https://github.com/lycosystem/lyscripts/compare/1.0.0rc3...1.0.0
 [1.0.0rc3]: https://github.com/lycosystem/lyscripts/compare/1.0.0rc2...1.0.0rc3
@@ -979,6 +990,7 @@ returns `None` instead. Fixes [#11]
 [#72]: https://github.com/lycosystem/lyscripts/issues/72
 [#74]: https://github.com/lycosystem/lyscripts/issues/74
 [#75]: https://github.com/lycosystem/lyscripts/issues/75
+[#81]: https://github.com/lycosystem/lyscripts/issues/81
 
 [`emcee`]: https://emcee.readthedocs.io/en/stable/
 [`rich`]: https://rich.readthedocs.io/en/latest/

pyproject.toml

@@ -10,7 +10,9 @@ build-backend = "setuptools.build_meta"
 name = "lyscripts"
 description = "Package to interact with lymphatic progression data and models."
 authors = [
-	{name = "Roman Ludwig", email = "roman.ludwig@usz.ch"}
+    {name = "Roman Ludwig", email = "gygqdstu3@mozmail.com"},
+    {name = "Yoel Pérez Haas", email = "yoel.perezhaas@usz.ch"},
+    {name = "Noemi Bührer", email = "noemi.buehrer@usz.ch"},
 ]
 readme = "README.md"
 requires-python = ">=3.10"

src/lyscripts/__init__.py

@@ -25,7 +25,7 @@
 __version__ = version
 __description__ = "Package to interact with lymphatic progression data and models."
 __author__ = "Roman Ludwig"
-__email__ = "roman.ludwig@usz.ch"
+__email__ = "gygqdstu3@mozmail.com"
 __uri__ = "https://github.com/lycosystem/lyscripts"
 
 # activate copy on write in pandas.

src/lyscripts/compute/evidence.py

@@ -48,23 +48,17 @@ def comp_bic(log_probs: np.ndarray, num_params: int, num_data: int) -> float:
 def compute_evidence(
     temp_schedule: np.ndarray,
     log_probs: np.ndarray,
-    num: int = 1000,
-) -> tuple[float, float]:
-    """Compute the evidence and its standard deviation.
+) -> float:
+    """Compute the evidence.
 
     Given a ``temp_schedule`` of inverse temperatures and corresponding sets of
-    ``log_probs``, draw ``num`` "paths" of log-probabilities and compute the evidence
-    for each using trapezoidal integration.
-
-    The evidence is then the mean of those ``num`` integrations, while the error is
-    their standard deviation.
+    ``log_probs``, we calculate the mean ``log_prob`` over all samples to approximate
+    the expectation value under the corresponding power posterior for each step in the
+    ``temp_schedule``. The evidence is evaluated using trapezoidal integration of the
+    expectation values over the ``temp_schedule``.
     """
-    integrals = np.zeros(shape=num)
-    for i in range(num):
-        rand_idx = RNG.choice(log_probs.shape[1], size=log_probs.shape[0])
-        drawn_accuracy = log_probs[np.arange(log_probs.shape[0]), rand_idx].copy()
-        integrals[i] = trapezoid(y=drawn_accuracy, x=temp_schedule)
-    return np.mean(integrals), np.std(integrals)
+    a_mc = np.mean(log_probs, axis=1)
+    return trapezoid(y=a_mc, x=temp_schedule)
 
 
 def compute_ti_results(
@@ -95,9 +89,8 @@ def compute_ti_results(
         )
         ti_log_probs[i] = reader.get_blobs(flat=True)["log_prob"]
 
-    evidence, evidence_std = compute_evidence(temp_schedule, ti_log_probs)
+    evidence = compute_evidence(temp_schedule, ti_log_probs)
     metrics["evidence"] = evidence
-    metrics["evidence_std"] = evidence_std
 
     return temp_schedule, ti_log_probs
 

src/lyscripts/evaluate.py

@@ -90,23 +90,17 @@ def comp_bic(log_probs: np.ndarray, num_params: int, num_data: int) -> float:
 def compute_evidence(
     temp_schedule: np.ndarray,
     log_probs: np.ndarray,
-    num: int = 1000,
-) -> tuple[float, float]:
-    """Compute the evidence and its standard deviation.
+) -> float:
+    """Compute the evidence.
 
     Given a ``temp_schedule`` of inverse temperatures and corresponding sets of
-    ``log_probs``, draw ``num`` "paths" of log-probabilities and compute the evidence
-    for each using trapezoidal integration.
-
-    The evidence is then the mean of those ``num`` integrations, while the error is
-    their standard deviation.
+    ``log_probs``, we calculate the mean ``log_prob`` over all samples to approximate
+    the expectation value under the corresponding power posterior for each step in the
+    ``temp_schedule``. The evidence is evaluated using trapezoidal integration of the
+    expectation values over the ``temp_schedule``.
     """
-    integrals = np.zeros(shape=num)
-    for i in range(num):
-        rand_idx = RNG.choice(log_probs.shape[1], size=log_probs.shape[0])
-        drawn_accuracy = log_probs[np.arange(log_probs.shape[0]), rand_idx].copy()
-        integrals[i] = trapezoid(y=drawn_accuracy, x=temp_schedule)
-    return np.mean(integrals), np.std(integrals)
+    a_mc = np.mean(log_probs, axis=1)
+    return trapezoid(y=a_mc, x=temp_schedule)
 
 
 def compute_ti_results(
@@ -134,9 +128,8 @@ def compute_ti_results(
         reader = emcee.backends.HDFBackend(model, name=f"ti/{run}", read_only=True)
         ti_log_probs[i] = reader.get_blobs(flat=True)
 
-    evidence, evidence_std = compute_evidence(temp_schedule, ti_log_probs)
+    evidence = compute_evidence(temp_schedule, ti_log_probs)
     metrics["evidence"] = evidence
-    metrics["evidence_std"] = evidence_std
 
     return temp_schedule, ti_log_probs