MaldiDeepKit 0.1.0__tar.gz → 0.2.0__tar.gz

{maldideepkit-0.1.0 → maldideepkit-0.2.0/MaldiDeepKit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: MaldiDeepKit
-Version: 0.1.0
+Version: 0.2.0
 Summary: A catalog of sklearn-compatible deep learning classifiers for MALDI-TOF binned spectra
 Author-email: Ettore Rocchi <ettore.rocchi3@unibo.it>
 License-Expression: MIT
@@ -45,6 +45,8 @@ Requires-Dist: sphinx-design>=0.6.0; extra == "docs"
 Requires-Dist: nbsphinx>=0.9.0; extra == "docs"
 Requires-Dist: ipykernel>=6.0.0; extra == "docs"
 Requires-Dist: ipywidgets>=8.0.0; extra == "docs"
+Provides-Extra: uncertainty
+Requires-Dist: laplace-torch<1.0,>=0.2; extra == "uncertainty"
 Dynamic: license-file
 
 # MaldiDeepKit
@@ -87,6 +89,24 @@ pip install maldideepkit
 
 `maldiamrkit` is a core dependency and is installed automatically - MaldiDeepKit duck-types on the `MaldiSet` data model and reuses `maldiamrkit.alignment.Warping` for leak-safe spectral warping.
 
+To enable the Laplace-approximation estimator in `maldideepkit.uncertainty`, install the optional `uncertainty` extra (pulls in [`laplace-torch`](https://github.com/aleximmer/Laplace)):
+
+```bash
+pip install "maldideepkit[uncertainty]"
+```
+
+Monte Carlo Dropout and split conformal prediction work without any extra.
+
+### Install the full MaldiSuite
+
+To install MaldiDeepKit together with [MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit) and [MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit) at compatible versions, install the [`maldisuite`](https://pypi.org/project/maldisuite/) meta-package:
+
+```bash
+pip install maldisuite
+```
+
+Visit the **MaldiSuite** landing page at <https://ettorerocchi.github.io/MaldiSuite/>.
+
 ### Development Installation
 
 ```bash
@@ -112,6 +132,7 @@ See [`CONTRIBUTING.md`](CONTRIBUTING.md) for coding conventions, testing, and PR
 - **Leak-safe spectral warping**: pass any sklearn-style transformer ([`maldiamrkit.alignment.Warping`](https://github.com/EttoreRocchi/MaldiAMRKit)) via `warping=`; it is fitted on the training fold only and applied to both splits during training and to new spectra at `predict` time, *before* per-feature standardization.
 - **MaldiSet integration**: pass a `maldiamrkit.MaldiSet` directly to `fit` / `predict`; MaldiDeepKit duck-types on the DataFrame-like `.X` attribute, so MaldiSuite's data model flows end-to-end.
 - **Persistence**: `save()` writes a state-dict `.pt` plus a hyperparameter `.json` (and a sibling `.warper.pkl` if a warper was fitted); `load()` fails fast on class or `input_dim` mismatches.
+- **Uncertainty quantification**: `maldideepkit.uncertainty` subpackage ships three drop-in estimators sharing a single `predict_with_uncertainty` interface. `MCDropoutEstimator` (Monte Carlo Dropout with epistemic / aleatoric decomposition), `LaplaceEstimator` (last-layer or full-network Laplace via the optional `laplace-torch` dependency), and `ConformalPredictor` (split conformal prediction with the LAC non-conformity score).
 - **CPU-friendly**: every classifier runs on CPU, which is what the project's CI tests against; CUDA speeds up the models' training significantly.
 
 ## Documentation
@@ -143,7 +164,7 @@ acc = clf.score(X, y)
 weights = clf.get_attention_weights(X[:10])  # (10, hidden_dim)
 ```
 
-### Inside an Sklearn Pipeline
+### Inside a scikit-learn `Pipeline`
 
 ```python
 from sklearn.model_selection import StratifiedKFold, cross_val_score
@@ -241,7 +262,7 @@ All four inherit from `BaseSpectralClassifier` and share the same hyperparameter
 
 - **`find_lr(clf, X, y)`** - learning-rate finder.
 - **`tune_threshold` / `fit_temperature`** - post-hoc calibrators usable standalone.
-- **`FocalLoss`, `SAMOptimizer`, `DropPath`** - building blocks for custom training loops.
+- **`FocalLoss`, `SAMOptimizer`** - loss / optimizer building blocks for custom training loops. Composable `nn.Module` primitives (`DropPath`, `PatchEmbed1D`, the full backbones) live under `maldideepkit.blocks`.
 
 ## Tutorials
 
@@ -251,16 +272,17 @@ For more detailed examples, see the notebooks:
 - [Model Comparison](notebooks/02_model_comparison.ipynb) - Train all four classifiers on the same dataset and compare accuracy.
 - [Attention Interpretation](notebooks/03_attention_interpretation.ipynb) - Visualise the sigmoid-gated attention learned by `MaldiMLPClassifier`.
 - [Full Pipeline](notebooks/04_full_pipeline.ipynb) - End-to-end template: MaldiAMRKit preprocessing + MaldiDeepKit classification.
+- [Uncertainty Quantification](notebooks/05_uncertainty.ipynb) - MC Dropout, split conformal prediction, and Laplace approximation on a fitted classifier; selective prediction curves.
 
 ## MaldiSuite Ecosystem
 
-MaldiDeepKit is the third package of the MaldiSuite ecosystem:
+MaldiDeepKit is the deep-learning package of the **MaldiSuite** ecosystem:
 
 - **[MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit)** - preprocessing, alignment, peak detection, differential analysis, and classical-ML evaluation for MALDI-TOF AMR workflows.
 - **[MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit)** - batch-effect correction and harmonisation for multi-centre / multi-instrument MALDI-TOF spectra.
 - **MaldiDeepKit** (this package) - sklearn-compatible deep learning classifiers.
 
-The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline.
+The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline. Install the full suite with `pip install maldisuite`. Landing page: [MaldiSuite](<https://ettorerocchi.github.io/MaldiSuite/>).
 
 ## Requirements
 
@@ -272,11 +294,9 @@ Pull requests, bug reports, and feature ideas are welcome. See the [Contributing
 
 ## Citing
 
-If you use MaldiDeepKit, please cite this repository until the companion paper is available.
-
-Related publications from the MaldiSuite ecosystem:
+If you use MaldiDeepKit in academic work please cite:
 
-> Rocchi, E., Nicitra, E., Calvo, M. et al. *Combining mass spectrometry and machine learning models for predicting Klebsiella pneumoniae antimicrobial resistance: a multicenter experience from clinical isolates in Italy*. **BMC Microbiol** (2026). [doi:10.1186/s12866-025-04657-2](https://link.springer.com/article/10.1186/s12866-025-04657-2)
+> _Citation will be available soon._
 
 See the [full publications list](https://maldideepkit.readthedocs.io/en/latest/papers.html) for more papers using the MaldiSuite.
 

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/MaldiDeepKit.egg-info/SOURCES.txt

@@ -3,6 +3,7 @@ README.md
 pyproject.toml
 requirements-dev.txt
 requirements-docs.txt
+requirements-uncertainty.txt
 requirements.txt
 MaldiDeepKit.egg-info/PKG-INFO
 MaldiDeepKit.egg-info/SOURCES.txt
@@ -28,6 +29,12 @@ maldideepkit/resnet/__init__.py
 maldideepkit/resnet/resnet.py
 maldideepkit/transformer/__init__.py
 maldideepkit/transformer/transformer.py
+maldideepkit/uncertainty/__init__.py
+maldideepkit/uncertainty/_base.py
+maldideepkit/uncertainty/_result.py
+maldideepkit/uncertainty/conformal.py
+maldideepkit/uncertainty/laplace.py
+maldideepkit/uncertainty/mc_dropout.py
 maldideepkit/utils/__init__.py
 maldideepkit/utils/calibration.py
 maldideepkit/utils/ensemble.py
@@ -51,4 +58,6 @@ tests/test_resnet.py
 tests/test_sam.py
 tests/test_sklearn_compat.py
 tests/test_transformer.py
+tests/test_uncertainty.py
+tests/test_warm_start.py
 tests/test_warping.py

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/MaldiDeepKit.egg-info/requires.txt

@@ -22,3 +22,6 @@ sphinx-design>=0.6.0
 nbsphinx>=0.9.0
 ipykernel>=6.0.0
 ipywidgets>=8.0.0
+
+[uncertainty]
+laplace-torch<1.0,>=0.2

{maldideepkit-0.1.0/MaldiDeepKit.egg-info → maldideepkit-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: MaldiDeepKit
-Version: 0.1.0
+Version: 0.2.0
 Summary: A catalog of sklearn-compatible deep learning classifiers for MALDI-TOF binned spectra
 Author-email: Ettore Rocchi <ettore.rocchi3@unibo.it>
 License-Expression: MIT
@@ -45,6 +45,8 @@ Requires-Dist: sphinx-design>=0.6.0; extra == "docs"
 Requires-Dist: nbsphinx>=0.9.0; extra == "docs"
 Requires-Dist: ipykernel>=6.0.0; extra == "docs"
 Requires-Dist: ipywidgets>=8.0.0; extra == "docs"
+Provides-Extra: uncertainty
+Requires-Dist: laplace-torch<1.0,>=0.2; extra == "uncertainty"
 Dynamic: license-file
 
 # MaldiDeepKit
@@ -87,6 +89,24 @@ pip install maldideepkit
 
 `maldiamrkit` is a core dependency and is installed automatically - MaldiDeepKit duck-types on the `MaldiSet` data model and reuses `maldiamrkit.alignment.Warping` for leak-safe spectral warping.
 
+To enable the Laplace-approximation estimator in `maldideepkit.uncertainty`, install the optional `uncertainty` extra (pulls in [`laplace-torch`](https://github.com/aleximmer/Laplace)):
+
+```bash
+pip install "maldideepkit[uncertainty]"
+```
+
+Monte Carlo Dropout and split conformal prediction work without any extra.
+
+### Install the full MaldiSuite
+
+To install MaldiDeepKit together with [MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit) and [MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit) at compatible versions, install the [`maldisuite`](https://pypi.org/project/maldisuite/) meta-package:
+
+```bash
+pip install maldisuite
+```
+
+Visit the **MaldiSuite** landing page at <https://ettorerocchi.github.io/MaldiSuite/>.
+
 ### Development Installation
 
 ```bash
@@ -112,6 +132,7 @@ See [`CONTRIBUTING.md`](CONTRIBUTING.md) for coding conventions, testing, and PR
 - **Leak-safe spectral warping**: pass any sklearn-style transformer ([`maldiamrkit.alignment.Warping`](https://github.com/EttoreRocchi/MaldiAMRKit)) via `warping=`; it is fitted on the training fold only and applied to both splits during training and to new spectra at `predict` time, *before* per-feature standardization.
 - **MaldiSet integration**: pass a `maldiamrkit.MaldiSet` directly to `fit` / `predict`; MaldiDeepKit duck-types on the DataFrame-like `.X` attribute, so MaldiSuite's data model flows end-to-end.
 - **Persistence**: `save()` writes a state-dict `.pt` plus a hyperparameter `.json` (and a sibling `.warper.pkl` if a warper was fitted); `load()` fails fast on class or `input_dim` mismatches.
+- **Uncertainty quantification**: `maldideepkit.uncertainty` subpackage ships three drop-in estimators sharing a single `predict_with_uncertainty` interface. `MCDropoutEstimator` (Monte Carlo Dropout with epistemic / aleatoric decomposition), `LaplaceEstimator` (last-layer or full-network Laplace via the optional `laplace-torch` dependency), and `ConformalPredictor` (split conformal prediction with the LAC non-conformity score).
 - **CPU-friendly**: every classifier runs on CPU, which is what the project's CI tests against; CUDA speeds up the models' training significantly.
 
 ## Documentation
@@ -143,7 +164,7 @@ acc = clf.score(X, y)
 weights = clf.get_attention_weights(X[:10])  # (10, hidden_dim)
 ```
 
-### Inside an Sklearn Pipeline
+### Inside a scikit-learn `Pipeline`
 
 ```python
 from sklearn.model_selection import StratifiedKFold, cross_val_score
@@ -241,7 +262,7 @@ All four inherit from `BaseSpectralClassifier` and share the same hyperparameter
 
 - **`find_lr(clf, X, y)`** - learning-rate finder.
 - **`tune_threshold` / `fit_temperature`** - post-hoc calibrators usable standalone.
-- **`FocalLoss`, `SAMOptimizer`, `DropPath`** - building blocks for custom training loops.
+- **`FocalLoss`, `SAMOptimizer`** - loss / optimizer building blocks for custom training loops. Composable `nn.Module` primitives (`DropPath`, `PatchEmbed1D`, the full backbones) live under `maldideepkit.blocks`.
 
 ## Tutorials
 
@@ -251,16 +272,17 @@ For more detailed examples, see the notebooks:
 - [Model Comparison](notebooks/02_model_comparison.ipynb) - Train all four classifiers on the same dataset and compare accuracy.
 - [Attention Interpretation](notebooks/03_attention_interpretation.ipynb) - Visualise the sigmoid-gated attention learned by `MaldiMLPClassifier`.
 - [Full Pipeline](notebooks/04_full_pipeline.ipynb) - End-to-end template: MaldiAMRKit preprocessing + MaldiDeepKit classification.
+- [Uncertainty Quantification](notebooks/05_uncertainty.ipynb) - MC Dropout, split conformal prediction, and Laplace approximation on a fitted classifier; selective prediction curves.
 
 ## MaldiSuite Ecosystem
 
-MaldiDeepKit is the third package of the MaldiSuite ecosystem:
+MaldiDeepKit is the deep-learning package of the **MaldiSuite** ecosystem:
 
 - **[MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit)** - preprocessing, alignment, peak detection, differential analysis, and classical-ML evaluation for MALDI-TOF AMR workflows.
 - **[MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit)** - batch-effect correction and harmonisation for multi-centre / multi-instrument MALDI-TOF spectra.
 - **MaldiDeepKit** (this package) - sklearn-compatible deep learning classifiers.
 
-The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline.
+The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline. Install the full suite with `pip install maldisuite`. Landing page: [MaldiSuite](<https://ettorerocchi.github.io/MaldiSuite/>).
 
 ## Requirements
 
@@ -272,11 +294,9 @@ Pull requests, bug reports, and feature ideas are welcome. See the [Contributing
 
 ## Citing
 
-If you use MaldiDeepKit, please cite this repository until the companion paper is available.
-
-Related publications from the MaldiSuite ecosystem:
+If you use MaldiDeepKit in academic work please cite:
 
-> Rocchi, E., Nicitra, E., Calvo, M. et al. *Combining mass spectrometry and machine learning models for predicting Klebsiella pneumoniae antimicrobial resistance: a multicenter experience from clinical isolates in Italy*. **BMC Microbiol** (2026). [doi:10.1186/s12866-025-04657-2](https://link.springer.com/article/10.1186/s12866-025-04657-2)
+> _Citation will be available soon._
 
 See the [full publications list](https://maldideepkit.readthedocs.io/en/latest/papers.html) for more papers using the MaldiSuite.
 

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/README.md

@@ -38,6 +38,24 @@ pip install maldideepkit
 
 `maldiamrkit` is a core dependency and is installed automatically - MaldiDeepKit duck-types on the `MaldiSet` data model and reuses `maldiamrkit.alignment.Warping` for leak-safe spectral warping.
 
+To enable the Laplace-approximation estimator in `maldideepkit.uncertainty`, install the optional `uncertainty` extra (pulls in [`laplace-torch`](https://github.com/aleximmer/Laplace)):
+
+```bash
+pip install "maldideepkit[uncertainty]"
+```
+
+Monte Carlo Dropout and split conformal prediction work without any extra.
+
+### Install the full MaldiSuite
+
+To install MaldiDeepKit together with [MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit) and [MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit) at compatible versions, install the [`maldisuite`](https://pypi.org/project/maldisuite/) meta-package:
+
+```bash
+pip install maldisuite
+```
+
+Visit the **MaldiSuite** landing page at <https://ettorerocchi.github.io/MaldiSuite/>.
+
 ### Development Installation
 
 ```bash
@@ -63,6 +81,7 @@ See [`CONTRIBUTING.md`](CONTRIBUTING.md) for coding conventions, testing, and PR
 - **Leak-safe spectral warping**: pass any sklearn-style transformer ([`maldiamrkit.alignment.Warping`](https://github.com/EttoreRocchi/MaldiAMRKit)) via `warping=`; it is fitted on the training fold only and applied to both splits during training and to new spectra at `predict` time, *before* per-feature standardization.
 - **MaldiSet integration**: pass a `maldiamrkit.MaldiSet` directly to `fit` / `predict`; MaldiDeepKit duck-types on the DataFrame-like `.X` attribute, so MaldiSuite's data model flows end-to-end.
 - **Persistence**: `save()` writes a state-dict `.pt` plus a hyperparameter `.json` (and a sibling `.warper.pkl` if a warper was fitted); `load()` fails fast on class or `input_dim` mismatches.
+- **Uncertainty quantification**: `maldideepkit.uncertainty` subpackage ships three drop-in estimators sharing a single `predict_with_uncertainty` interface. `MCDropoutEstimator` (Monte Carlo Dropout with epistemic / aleatoric decomposition), `LaplaceEstimator` (last-layer or full-network Laplace via the optional `laplace-torch` dependency), and `ConformalPredictor` (split conformal prediction with the LAC non-conformity score).
 - **CPU-friendly**: every classifier runs on CPU, which is what the project's CI tests against; CUDA speeds up the models' training significantly.
 
 ## Documentation
@@ -94,7 +113,7 @@ acc = clf.score(X, y)
 weights = clf.get_attention_weights(X[:10])  # (10, hidden_dim)
 ```
 
-### Inside an Sklearn Pipeline
+### Inside a scikit-learn `Pipeline`
 
 ```python
 from sklearn.model_selection import StratifiedKFold, cross_val_score
@@ -192,7 +211,7 @@ All four inherit from `BaseSpectralClassifier` and share the same hyperparameter
 
 - **`find_lr(clf, X, y)`** - learning-rate finder.
 - **`tune_threshold` / `fit_temperature`** - post-hoc calibrators usable standalone.
-- **`FocalLoss`, `SAMOptimizer`, `DropPath`** - building blocks for custom training loops.
+- **`FocalLoss`, `SAMOptimizer`** - loss / optimizer building blocks for custom training loops. Composable `nn.Module` primitives (`DropPath`, `PatchEmbed1D`, the full backbones) live under `maldideepkit.blocks`.
 
 ## Tutorials
 
@@ -202,16 +221,17 @@ For more detailed examples, see the notebooks:
 - [Model Comparison](notebooks/02_model_comparison.ipynb) - Train all four classifiers on the same dataset and compare accuracy.
 - [Attention Interpretation](notebooks/03_attention_interpretation.ipynb) - Visualise the sigmoid-gated attention learned by `MaldiMLPClassifier`.
 - [Full Pipeline](notebooks/04_full_pipeline.ipynb) - End-to-end template: MaldiAMRKit preprocessing + MaldiDeepKit classification.
+- [Uncertainty Quantification](notebooks/05_uncertainty.ipynb) - MC Dropout, split conformal prediction, and Laplace approximation on a fitted classifier; selective prediction curves.
 
 ## MaldiSuite Ecosystem
 
-MaldiDeepKit is the third package of the MaldiSuite ecosystem:
+MaldiDeepKit is the deep-learning package of the **MaldiSuite** ecosystem:
 
 - **[MaldiAMRKit](https://github.com/EttoreRocchi/MaldiAMRKit)** - preprocessing, alignment, peak detection, differential analysis, and classical-ML evaluation for MALDI-TOF AMR workflows.
 - **[MaldiBatchKit](https://github.com/EttoreRocchi/MaldiBatchKit)** - batch-effect correction and harmonisation for multi-centre / multi-instrument MALDI-TOF spectra.
 - **MaldiDeepKit** (this package) - sklearn-compatible deep learning classifiers.
 
-The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline.
+The three packages share the `MaldiSet` / `MaldiSpectrum` data model and are designed to compose in a single end-to-end pipeline. Install the full suite with `pip install maldisuite`. Landing page: [MaldiSuite](<https://ettorerocchi.github.io/MaldiSuite/>).
 
 ## Requirements
 
@@ -223,11 +243,9 @@ Pull requests, bug reports, and feature ideas are welcome. See the [Contributing
 
 ## Citing
 
-If you use MaldiDeepKit, please cite this repository until the companion paper is available.
-
-Related publications from the MaldiSuite ecosystem:
+If you use MaldiDeepKit in academic work please cite:
 
-> Rocchi, E., Nicitra, E., Calvo, M. et al. *Combining mass spectrometry and machine learning models for predicting Klebsiella pneumoniae antimicrobial resistance: a multicenter experience from clinical isolates in Italy*. **BMC Microbiol** (2026). [doi:10.1186/s12866-025-04657-2](https://link.springer.com/article/10.1186/s12866-025-04657-2)
+> _Citation will be available soon._
 
 See the [full publications list](https://maldideepkit.readthedocs.io/en/latest/papers.html) for more papers using the MaldiSuite.
 

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/__init__.py

@@ -1,4 +1,4 @@
-"""MaldiDeepKit -- deep learning classifiers for MALDI-TOF binned spectra.
+"""MaldiDeepKit - deep learning classifiers for MALDI-TOF binned spectra.
 
 Provides a catalog of PyTorch architectures (MLP, CNN, ResNet,
 Transformer) adapted to 1-D binned MALDI-TOF spectra, each wrapped in
@@ -6,18 +6,21 @@ a scikit-learn compatible estimator with sensible defaults.
 
 Subpackages
 -----------
-- ``maldideepkit.base`` -- ``BaseSpectralClassifier``, ``SpectralDataset``,
+- ``maldideepkit.base`` - ``BaseSpectralClassifier``, ``SpectralDataset``,
   ``make_loaders``.
-- ``maldideepkit.attention`` -- ``MaldiMLPClassifier`` (MLP with optional
+- ``maldideepkit.attention`` - ``MaldiMLPClassifier`` (MLP with optional
   sigmoid-gated attention).
-- ``maldideepkit.cnn`` -- ``MaldiCNNClassifier`` (Conv1D blocks).
-- ``maldideepkit.resnet`` -- ``MaldiResNetClassifier`` (1-D ResNet-18).
-- ``maldideepkit.transformer`` -- ``MaldiTransformerClassifier`` (1-D ViT).
-- ``maldideepkit.blocks`` -- re-exports of every backbone and
+- ``maldideepkit.cnn`` - ``MaldiCNNClassifier`` (Conv1D blocks).
+- ``maldideepkit.resnet`` - ``MaldiResNetClassifier`` (1-D ResNet-18).
+- ``maldideepkit.transformer`` - ``MaldiTransformerClassifier`` (1-D ViT).
+- ``maldideepkit.blocks`` - re-exports of every backbone and
   composable primitive for users embedding components into their own
   networks.
-- ``maldideepkit.utils`` -- reproducibility helpers and shared
+- ``maldideepkit.utils`` - reproducibility helpers and shared
   training primitives.
+- ``maldideepkit.uncertainty`` - uncertainty-quantification
+  estimators (MC Dropout, Laplace approximation, split conformal
+  prediction) for fitted classifiers.
 
 Examples
 --------
@@ -31,6 +34,7 @@ Examples
 >>> proba = clf.predict_proba(X)
 """
 
+from . import uncertainty
 from .attention.mlp import MaldiMLPClassifier
 from .base.classifier import BaseSpectralClassifier
 from .base.data import SpectralDataset, make_loaders
@@ -38,7 +42,7 @@ from .cnn.cnn import MaldiCNNClassifier
 from .resnet.resnet import MaldiResNetClassifier
 from .transformer.transformer import MaldiTransformerClassifier
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"
 __author__ = "Ettore Rocchi"
 
 __all__ = [
@@ -51,4 +55,5 @@ __all__ = [
     "__author__",
     "__version__",
     "make_loaders",
+    "uncertainty",
 ]

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/attention/mlp.py

@@ -121,12 +121,14 @@ class MaldiMLPClassifier(BaseSpectralClassifier):
         Dropout after the projection and first head layer.
     dropout_low : float, default=0.2
         Dropout before the output logits.
-    **kwargs
-        Forwarded to :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`:
-        ``input_dim``, ``n_classes``, ``learning_rate``, ``batch_size``,
-        ``epochs``, ``early_stopping_patience``, ``val_fraction``,
-        ``standardize``, ``class_weight``, ``device``, ``random_state``,
-        ``verbose``.
+
+    Notes
+    -----
+    Every parameter accepted by
+    :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`
+    (e.g. ``learning_rate``, ``batch_size``, ``epochs``, ``warping``,
+    ``calibrate_temperature``, ``device``, ``random_state``, ...) is
+    forwarded to the base class. See its docstring for the full list.
 
     Attributes
     ----------
@@ -258,12 +260,14 @@ class MaldiMLPClassifier(BaseSpectralClassifier):
             self.attention_weights_ = None
         return logits
 
-    def fit(self, X: Any, y: Any) -> MaldiMLPClassifier:  # type: ignore[override]
+    def fit(  # type: ignore[override]
+        self, X: Any, y: Any, *, warm_start: bool = False
+    ) -> MaldiMLPClassifier:
         """Fit the model and cache attention weights from the final batch.
 
-        See :meth:`BaseSpectralClassifier.fit` for shared parameters.
+        See :meth:`BaseSpectralClassifier.fit` for shared parameters, including ``warm_start``.
         """
-        super().fit(X, y)
+        super().fit(X, y, warm_start=warm_start)
         if self.use_attention:
             X_np = _to_numpy(X)
             tail = X_np[: min(len(X_np), 64)]

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/base/__init__.py

@@ -1,8 +1,9 @@
 """Shared primitives for MaldiDeepKit classifiers.
 
-Exposes :class:`BaseSpectralClassifier`, the abstract base for all six
-model families in the package, together with the :class:`SpectralDataset`
-/ :func:`make_loaders` data utilities.
+Exposes :class:`BaseSpectralClassifier`, the abstract base for the four
+classifier families in the package (MLP, CNN, ResNet, Transformer),
+together with the :class:`SpectralDataset` / :func:`make_loaders` data
+utilities.
 """
 
 from .classifier import BaseSpectralClassifier

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/base/classifier.py

@@ -399,7 +399,9 @@ class BaseSpectralClassifier(ClassifierMixin, BaseEstimator, metaclass=ABCMeta):
         y_encoded = np.searchsorted(self.classes_, y_np).astype(np.int64)
         return X_np, y_encoded
 
-    def fit(self, X: Any, y: Any) -> BaseSpectralClassifier:
+    def fit(
+        self, X: Any, y: Any, *, warm_start: bool = False
+    ) -> BaseSpectralClassifier:
         """Fit the model on ``(X, y)``.
 
         Parameters
@@ -410,6 +412,20 @@ class BaseSpectralClassifier(ClassifierMixin, BaseEstimator, metaclass=ABCMeta):
         y : array-like of shape (n_samples,)
             Integer or string class labels. Re-encoded to ``0..n_classes-1``
             internally; original labels are preserved in :attr:`classes_`.
+        warm_start : bool, default=False
+            When ``True`` and the estimator already has a fitted
+            :attr:`model_`, the underlying :class:`torch.nn.Module` is
+            reused as the starting point of training instead of being
+            rebuilt from scratch via :meth:`_build_model`. This unblocks
+            federated learning, continual learning, and fine-tuning
+            workflows that need ``fit()`` to *resume* from the current
+            weights rather than reinitialise. ``warm_start`` applies only
+            to the first training attempt; retries triggered by
+            ``retry_on_val_auroc_below`` always rebuild via
+            :meth:`_build_model` (the warm-start weights already failed
+            once). When ``warm_start=True`` but no prior ``model_``
+            exists, falls back silently to a fresh build (sklearn
+            convention).
 
         Returns
         -------
@@ -456,7 +472,17 @@ class BaseSpectralClassifier(ClassifierMixin, BaseEstimator, metaclass=ABCMeta):
             if attempt > 0:
                 seed_everything(base_seed + 1_000_003 * attempt)
 
-            model = self._build_model().to(device)
+            if (
+                warm_start
+                and attempt == 0
+                and getattr(self, "model_", None) is not None
+            ):
+                # Resume from the previously-fitted module. Skip
+                # _build_model entirely so the federated / continual
+                # learning caller's pre-loaded weights are not wiped.
+                model = self.model_.to(device)
+            else:
+                model = self._build_model().to(device)
 
             opt_cls = (
                 torch.optim.AdamW if float(self.weight_decay) > 0 else torch.optim.Adam

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/cnn/cnn.py

@@ -150,11 +150,15 @@ class MaldiCNNClassifier(BaseSpectralClassifier):
         Width of the hidden dense layer.
     dropout : float, default=0.3
         Dropout applied inside every block and before the output layer.
-    **kwargs
-        Forwarded to :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`.
 
     Notes
     -----
+    Every parameter accepted by
+    :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`
+    (e.g. ``learning_rate``, ``batch_size``, ``epochs``, ``warping``,
+    ``calibrate_temperature``, ``device``, ``random_state``, ...) is
+    forwarded to the base class. See its docstring for the full list.
+
     The flat dense head scales linearly with ``input_dim``; prefer
     :class:`~maldideepkit.MaldiResNetClassifier` or
     :class:`~maldideepkit.MaldiTransformerClassifier` if you want

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/resnet/resnet.py

@@ -35,8 +35,6 @@ class BasicBlock1D(nn.Module):
         Kernel size of both Conv1D layers.
     """
 
-    expansion = 1
-
     def __init__(
         self,
         in_channels: int,
@@ -220,11 +218,18 @@ class MaldiResNetClassifier(BaseSpectralClassifier):
         reproduce the literal ResNet-18 backbone.
     dropout : float, default=0.2
         Dropout before the final linear layer.
-    **kwargs
-        Forwarded to :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`.
 
     Notes
     -----
+    Every parameter accepted by
+    :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`
+    (e.g. ``learning_rate``, ``batch_size``, ``epochs``, ``warping``,
+    ``calibrate_temperature``, ``device``, ``random_state``, ...) is
+    forwarded to the base class. The ResNet defaults pre-set
+    ``weight_decay=1e-4``, ``grad_clip_norm=1.0``, ``warmup_epochs=5``,
+    and ``input_transform="log1p"``; override any of them at
+    construction time.
+
     Defaults deviate from literal ResNet-18 (He et al., 2016) in three
     ways for MALDI-TOF: ``stem_stride=1`` (was 2),
     ``block_kernel_size=7`` (was 3), and ``use_stem_pool=False`` (was

{maldideepkit-0.1.0 → maldideepkit-0.2.0}/maldideepkit/transformer/transformer.py

@@ -336,11 +336,15 @@ class MaldiTransformerClassifier(BaseSpectralClassifier):
         Token aggregation for classification.
     head_dim : int, default=128
         Width of the hidden dense layer in the classification head.
-    **kwargs
-        Forwarded to :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`.
 
     Notes
     -----
+    Every parameter accepted by
+    :class:`~maldideepkit.base.classifier.BaseSpectralClassifier`
+    (e.g. ``learning_rate``, ``batch_size``, ``epochs``, ``warping``,
+    ``calibrate_temperature``, ``device``, ``random_state``, ...) is
+    forwarded to the base class. See its docstring for the full list.
+
     Transformer training recipe baked in as defaults: ``lr=3e-4``,
     ``weight_decay=0.05``, ``grad_clip_norm=1.0``, ``warmup_epochs=5``.
 

maldideepkit-0.2.0/maldideepkit/uncertainty/__init__.py

@@ -0,0 +1,34 @@
+"""Uncertainty-quantification estimators for fitted MaldiDeepKit classifiers.
+
+Three drop-in estimators that share a single
+:meth:`predict_with_uncertainty` interface:
+
+- :class:`MCDropoutEstimator` - Monte Carlo Dropout. Stochastic
+  forward passes with dropout active; decomposes total uncertainty
+  into epistemic (model disagreement) and aleatoric (data noise)
+  components.
+- :class:`LaplaceEstimator` - Last-layer or full-network Laplace
+  approximation via the optional ``laplace-torch`` dependency.
+- :class:`ConformalPredictor` - Distribution-free split conformal
+  prediction with the LAC non-conformity score; produces calibrated
+  prediction sets without retraining the model.
+
+All three return a :class:`UncertaintyResult` so downstream code can
+swap methods without changing call sites.
+"""
+
+from __future__ import annotations
+
+from ._base import BaseUncertaintyEstimator
+from ._result import UncertaintyResult
+from .conformal import ConformalPredictor
+from .laplace import LaplaceEstimator
+from .mc_dropout import MCDropoutEstimator
+
+__all__ = [
+    "BaseUncertaintyEstimator",
+    "ConformalPredictor",
+    "LaplaceEstimator",
+    "MCDropoutEstimator",
+    "UncertaintyResult",
+]