Added hyperparameter optimization and updated README

README.md

@@ -39,7 +39,7 @@ pip install --upgrade qudi-hira-analysis
 
 If you are publishing scientific results, you can cite this work as:  https://doi.org/10.5281/zenodo.7604670
 
-## Examples
+## Usage
 
 First set up the `DataHandler` object (henceforth referred to as `dh`) with the correct paths to the data and figure
 folders.
@@ -89,7 +89,56 @@ The `load_measurements` function returns a dictionary containing the measurement
 - The dictionary values are `MeasurementDataclass` objects whose schema is shown
   visually [here](#measurement-dataclass-schema).
 
-### Example 0: NV-PL measurements
+### Example 0: 2D NV-ODMR measurements
+
+```python
+odmr_measurements = dh.load_measurements(measurement_str="2d_odmr_map")
+odmr_measurements = dict(sorted(odmr_measurements.items()))
+
+# Optional: Try and optimize the hyperparameters for the ODMR fitting
+highest_min_r2, optimal_parameters = dh.optimize_hyperparameters(odmr_measurements, num_samples=100, num_params=3)
+
+# Perform parallel (=num CPU cores) ODMR fitting
+odmr_measurements = dh.raster_odmr_fitting(
+    odmr_measurements,
+    r2_thresh=0.95,
+    thresh_frac=0.5,
+    sigma_thresh_frac=0.1,
+    min_thresh=0.01,
+)
+
+# Calculate residuals and 2D ODMR map
+pixels = int(np.sqrt(len(odmr_measurements)))
+image = np.zeros((pixels, pixels))
+residuals = np.zeros(len(odmr_measurements))
+
+for idx, odmr in enumerate(odmr_measurements.values()):
+    row, col = odmr.xy_position
+    residuals[idx] = odmr.fit_model.rsquared
+
+    if len(odmr.fit_model.params) == 6:
+        # Single Lorentzian, no splitting
+        image[row, col] = 0
+    else:
+        if odmr.fit_model.rsquared < 0.95:
+            # Bad fit, set to NaN
+            image[row, col] = np.nan
+        else:
+            # Calculate splitting
+            splitting = np.abs(odmr.fit_model.best_values["l1_center"] - odmr.fit_model.best_values["l0_center"])
+            image[row, col] = splitting
+
+fig, (ax, ax1) = plt.subplots(ncols=2)
+# Plot residuals
+sns.lineplot(residuals, ax=ax)
+# Plot 2D ODMR map
+sns.heatmap(image, cbar_kws={"label": r"$\Delta E$ (MHz)"}, ax=ax1)
+
+# Save the figure to the figure folder specified earlier
+dh.save_figures(filepath="2d_odmr_map_with_residuals", fig=fig, only_jpg=True)
+```
+
+### Example 1: NV-PL measurements
 
 ```python
 pixel_scanner_measurements = dh.load_measurements(measurement_str="PixelScanner")
@@ -114,7 +163,7 @@ cbar.set_label("NV-PL (kcps)")
 dh.save_figures(filepath="nv_pl_scan", fig=fig, only_jpg=True)
 ```
 
-### Example 1: Nanonis AFM measurements
+### Example 2: Nanonis AFM measurements
 
 ```python
 afm_measurements = dh.load_measurements(measurement_str="Scan", extension=".sxm", qudi=False)
@@ -142,7 +191,7 @@ cbar.set_label("Height (nm)")
 dh.save_figures(filepath="afm_topo", fig=fig, only_jpg=True)
 ``` 
 
-### Example 2: Autocorrelation measurements (Antibunching fit)
+### Example 3: Autocorrelation measurements (Antibunching fit)
 
 ```python
 autocorrelation_measurements = dh.load_measurements(measurement_str="Autocorrelation")
@@ -163,7 +212,7 @@ for autocorrelation in autocorrelation_measurements.values():
 dh.save_figures(filepath="autocorrelation_variation", fig=fig)
 ```
 
-### Example 3: ODMR measurements (double Lorentzian fit)
+### Example 4: ODMR measurements (double Lorentzian fit)
 
 ```python
 odmr_measurements = dh.load_measurements(measurement_str="ODMR", pulsed=True)
@@ -179,7 +228,7 @@ for odmr in odmr_measurements.values():
 dh.save_figures(filepath="odmr_variation", fig=fig)
 ```
 
-### Example 4: Rabi measurements (sine exponential decay fit)
+### Example 5: Rabi measurements (sine exponential decay fit)
 
 ```python
 rabi_measurements = dh.load_measurements(measurement_str="Rabi", pulsed=True)
@@ -195,7 +244,7 @@ for rabi in rabi_measurements.values():
 dh.save_figures(filepath="rabi_variation", fig=fig)
 ```
 
-### Example 5: Temperature data
+### Example 6: Temperature data
 
 ```python
 temperature_measurements = dh.load_measurements(measurement_str="Temperature", qudi=False)
@@ -207,7 +256,7 @@ sns.lineplot(data=temperature, x="Time", y="Temperature", ax=ax)
 dh.save_figures(filepath="temperature_monitoring", fig=fig)
 ```
 
-### Example 6: PYS data (pi3diamond compatibility)
+### Example 7: PYS data (pi3diamond compatibility)
 
 ```python
 pys_measurements = dh.load_measurements(measurement_str="ndmin", extension=".pys", qudi=False)
@@ -218,7 +267,7 @@ sns.lineplot(x=pys["time_bins"], y=pys["counts"], ax=ax)
 dh.save_figures(filepath="pys_measurement", fig=fig)
 ```
 
-### Example 7: Bruker MFM data
+### Example 8: Bruker MFM data
 
 ```python
 bruker_measurements = dh.load_measurements(measurement_str="", extension=".001", qudi=False)

qudi_hira_analysis/analysis_logic.py

@@ -1,7 +1,9 @@
 from __future__ import annotations
 
 import logging
+import random
 import re
+from itertools import product
 from typing import Tuple, TYPE_CHECKING
 
 import numpy as np
@@ -287,13 +289,65 @@ def analyse_mean_norm(
 
         return signal_data, error_data
 
-    def raster_odmr_fitting(
+    def optimize_hyperparameters(
             self,
+            measurements: dict[str, MeasurementDataclass],
+            num_samples: int = 10,
+            num_params: int = 3,
+    ) -> Tuple[float, Tuple[float, float, float]]:
+        """
+        This method optimizes the hyperparameters of the ODMR analysis.
+        It does so by randomly sampling a subset of the measurements and
+        then optimizing the hyperparameters for them.
+
+        Args:
+            measurements: A dictionary of measurements to optimize the hyperparameters for.
+            num_params: The number of parameters to optimize.
+            num_samples: The number of measurements to sample.
+
+        Returns:
+            The optimal hyperparameters.
+        """
+        r2_threshs = np.around(np.linspace(start=0.9, stop=0.99, num=num_params), decimals=2)
+        thresh_fracs = np.around(np.linspace(start=0.5, stop=0.9, num=num_params), decimals=1)
+        sigma_thresh_fracs = np.around(np.linspace(start=0.1, stop=0.2, num=num_params), decimals=1)
+
+        odmr_sample = {}
+        for k, v in random.sample(sorted(measurements.items()), k=num_samples):
+            odmr_sample[k] = v
+
+        highest_min_r2 = 0
+        optimal_params = (0, 0, 0)
+
+        for idx, (r2_thresh, thresh_frac, sigma_thresh_frac) in enumerate(
+                product(r2_threshs, thresh_fracs, sigma_thresh_fracs)):
+            odmr_sample = self.raster_odmr_fitting(
+                odmr_sample,
+                r2_thresh=r2_thresh,
+                thresh_frac=thresh_frac,
+                sigma_thresh_frac=sigma_thresh_frac,
+                min_thresh=0.01,
+                progress_bar=False
+            )
+
+            r2s = np.zeros(len(odmr_sample))
+            for _idx, odmr in enumerate(odmr_sample.values()):
+                r2s[_idx] = odmr.fit_model.rsquared
+            min_r2 = np.min(r2s)
+
+            if highest_min_r2 < min_r2:
+                highest_min_r2 = min_r2
+                optimal_params = (r2_thresh, thresh_frac, sigma_thresh_frac)
+
+        return highest_min_r2, optimal_params
+
+    @staticmethod
+    def raster_odmr_fitting(
             odmr_measurements: dict[str, MeasurementDataclass],
             r2_thresh: float = 0.95,
-            thresh_frac: float = 0.3,
-            min_thresh: float = 0.25,
-            sigma_thresh_frac: float = 0.3,
+            thresh_frac: float = 0.5,
+            sigma_thresh_frac: float = 0.15,
+            min_thresh: float = 0.01,
             extract_pixel_from_filename: bool = True,
             progress_bar: bool = True
     ) -> dict[str, MeasurementDataclass]:
@@ -307,7 +361,7 @@ def raster_odmr_fitting(
             min_thresh:
             sigma_thresh_frac:
             extract_pixel_from_filename: Extract `(row, col)` (in this format) from filename
-
+            progress_bar: Show progress bar
         Returns:
             List of ODMR data with fit, fit model and pixels in MeasurementDataclass
         """