update tutorials for code updaets

TomDonoghue · TomDonoghue · commit f18442df6ce7 · 2025-04-09T17:34:17.000-04:00
diff --git a/tutorials/plot_01-ModelDescription.py b/tutorials/plot_01-ModelDescription.py
@@ -50,10 +50,10 @@
 set_random_seed(21)
 
 # Simulate example power spectra
-freqs1, powers1 = sim_power_spectrum([3, 40], [1, 1],
-                                     [[10, 0.2, 1.25], [30, 0.15, 2]])
-freqs2, powers2 = sim_power_spectrum([1, 150], [1, 125, 1.25],
-                                     [[8, 0.15, 1.], [30, 0.1, 2]])
+freqs1, powers1 = sim_power_spectrum([3, 40], {'fixed' : [1, 1]},
+                                     {'gaussian' : [[10, 0.2, 1.25], [30, 0.15, 2]]})
+freqs2, powers2 = sim_power_spectrum([1, 150], {'knee' : [1, 125, 1.25]},
+                                     {'gaussian' : [[8, 0.15, 1.], [30, 0.1, 2]]})
 
 ###################################################################################################
 
diff --git a/tutorials/plot_02-PSDModel.py b/tutorials/plot_02-PSDModel.py
@@ -125,8 +125,8 @@
 
 # Goodness of fit measures
 print('Goodness of fit:')
-print(' Error - ', fm.results.error_)
-print(' R^2   - ', fm.results.r_squared_, '\n')
+print(' Error - ', fm.results.metrics.results['error_mae'])
+print(' R^2   - ', fm.results.metrics.results['gof_rsquared'], '\n')
 
 # Check how many peaks were fit
 print('Number of fit peaks: \n', fm.results.n_peaks_)
@@ -142,7 +142,7 @@
 ###################################################################################################
 
 # Extract a model parameter with `get_params`
-err = fm.get_params('error')
+err = fm.get_params('metrics', 'error_mae')
 
 # Extract parameters, indicating sub-selections of parameters
 exp = fm.get_params('aperiodic_params', 'exponent')
@@ -227,7 +227,7 @@
 fres = fm.results.get_results()
 
 # You can also unpack all fit parameters when using `get_results`
-ap_params, peak_params, r_squared, fit_error, gauss_params = fm.results.get_results()
+ap_params, peak_params, metrics, gauss_params = fm.results.get_results()
 
 ###################################################################################################
 
@@ -238,8 +238,8 @@
 print('Aperiodic Parameters: \n', fres.aperiodic_params)
 
 # Check the R^2 and error of the model fit
-print('R-squared: \n {:5.4f}'.format(fres.r_squared))
-print('Fit error: \n {:5.4f}'.format(fres.error))
+print('R-squared: \n {:5.4f}'.format(fres.metrics['gof_rsquared']))
+print('Fit error: \n {:5.4f}'.format(fres.metrics['error_mae']))
 
 ###################################################################################################
 # Conclusion
diff --git a/tutorials/plot_03-Algorithm.py b/tutorials/plot_03-Algorithm.py
@@ -107,7 +107,8 @@
 # Do an initial aperiodic fit - a robust fit, that excludes outliers
 #   This recreates an initial fit that isn't ultimately stored in the model object
 init_ap_fit = gen_aperiodic(\
-    fm.data.freqs, fm.algorithm._robust_ap_fit(fm.data.freqs, fm.data.power_spectrum))
+    fm.data.freqs, fm.modes.aperiodic,
+    fm.algorithm._robust_ap_fit(fm.data.freqs, fm.data.power_spectrum))
 
 # Plot the initial aperiodic fit
 _, ax = plt.subplots(figsize=(12, 10))
diff --git a/tutorials/plot_04-ModelObject.py b/tutorials/plot_04-ModelObject.py
@@ -230,12 +230,13 @@
 
 ###################################################################################################
 
-# Print out model fit results
+# Print out model fit results parameters
 print('aperiodic params: \t', fm.results.aperiodic_params_)
 print('peak params: \t', fm.results.peak_params_)
-print('r-squared: \t', fm.results.r_squared_)
-print('fit error: \t', fm.results.error_)
-print('modeled spectrum: \t', fm.results.modeled_spectrum_[0:5])
+
+# Print out metrics model fit results parameters
+print('fit error: \t', fm.results.metrics.results['error_mae'])
+print('r-squared: \t', fm.results.metrics.results['gof_rsquared'])
 
 ###################################################################################################
 # 4) Methods
diff --git a/tutorials/plot_06-GroupFits.py b/tutorials/plot_06-GroupFits.py
@@ -149,8 +149,8 @@
 cfs = fg.get_params('peak_params', 'CF')
 
 # Extract goodness-of-fit metrics
-errors = fg.get_params('error')
-r2s = fg.get_params('r_squared')
+errors = fg.get_params('metrics', 'error_mae')
+r2s = fg.get_params('metrics', 'gof_rsquared')
 
 ###################################################################################################
 
diff --git a/tutorials/plot_07-TimeModels.py b/tutorials/plot_07-TimeModels.py
@@ -19,7 +19,6 @@
 from specparam.sim import sim_spectrogram
 from specparam.plts.spectra import plot_spectrogram
 
-
 ###################################################################################################
 # Parameterizing Spectrograms
 # ---------------------------
@@ -42,38 +41,20 @@
 
 ###################################################################################################
 
-# Create & plot an example spectrogram
+# Define simulation parameters for a spectrogram
 n_pre_post = 50
 freq_range = [3, 25]
-ap_params = [[1, 1.5]] * n_pre_post + [[1, 1]] * n_pre_post
-pe_params = [[10, 1.5, 2.5]] * n_pre_post + [[10, 0.5, 2.]] * n_pre_post
+ap_params = {'fixed' : [[1, 1.5]] * n_pre_post + [[1, 1]] * n_pre_post}
+pe_params = {'gaussian' : [[10, 1.5, 2.5]] * n_pre_post + [[10, 0.5, 2.]] * n_pre_post}
+
+# Simulate spectrogram
 freqs, spectrogram = sim_spectrogram(n_pre_post * 2, freq_range, ap_params, pe_params, nlvs=0.1)
 
 ###################################################################################################
 
 # Plot our simulated spectrogram
 plot_spectrogram(freqs, spectrogram)
 
-###################################################################################################
-# SpectralTimeModel
-# -----------------
-#
-# The :class:`~specparam.SpectralTimeModel` object is an extension of the SpectralModel objects
-# to support parameterizing neural power spectra that are organized across time (spectrograms).
-#
-# In practice, this object is very similar to the previously introduced spectral model objects,
-# especially the Group model object. The time object is a mildly updated Group object.
-#
-# The main differences with the SpectralTimeModel from previous model objects are that the
-# data it accepts and parameterizes should be organized as as array of power spectra over
-# time windows - basically as a spectrogram.
-#
-
-###################################################################################################
-
-# Initialize a SpectralTimeModel model, which accepts all the same settings as SpectralModel
-ft = SpectralTimeModel()
-
 ###################################################################################################
 # Defining Oscillation Bands
 # ~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -93,6 +74,26 @@
 # Define a bands object to organize peak parameters
 bands = Bands({'alpha' : [7, 14]})
 
+###################################################################################################
+# SpectralTimeModel
+# -----------------
+#
+# The :class:`~specparam.SpectralTimeModel` object is an extension of the SpectralModel objects
+# to support parameterizing neural power spectra that are organized across time (spectrograms).
+#
+# In practice, this object is very similar to the previously introduced spectral model objects,
+# especially the Group model object. The time object is a mildly updated Group object.
+#
+# The main differences with the SpectralTimeModel from previous model objects are that the
+# data it accepts and parameterizes should be organized as as array of power spectra over
+# time windows - basically as a spectrogram.
+#
+
+###################################################################################################
+
+# Initialize a SpectralTimeModel model, which accepts all the same settings as SpectralModel
+ft = SpectralTimeModel(bands=bands)
+
 ###################################################################################################
 #
 # Now we are ready to fit our spectrogram! As with all model objects, we can fit the models
@@ -102,7 +103,7 @@
 ###################################################################################################
 
 # Fit the spectrogram and print out report
-ft.report(freqs, spectrogram, peak_org=bands)
+ft.report(freqs, spectrogram)
 
 ###################################################################################################
 #
@@ -129,7 +130,8 @@
 n_events = 3
 spectrograms = []
 for ind in range(n_events):
-    freqs, cur_spect = sim_spectrogram(n_pre_post * 2, freq_range, ap_params, pe_params, nlvs=0.1)
+    freqs, cur_spect = sim_spectrogram(\
+        n_pre_post * 2, freq_range, ap_params, pe_params, nlvs=0.1)
     spectrograms.append(cur_spect)
 
 ###################################################################################################
@@ -157,12 +159,12 @@
 ###################################################################################################
 
 # Initialize the spectral event model
-fe = SpectralTimeEventModel()
+fe = SpectralTimeEventModel(bands=bands)
 
 ###################################################################################################
 
 # Fit the spectrograms and print out report
-fe.report(freqs, spectrograms, peak_org=bands)
+fe.report(freqs, spectrograms)
 
 ###################################################################################################
 #
diff --git a/tutorials/plot_08-TroubleShooting.py b/tutorials/plot_08-TroubleShooting.py
@@ -137,7 +137,8 @@
 set_random_seed(21)
 
 # Create a simulated power spectrum
-freqs, spectrum = sim_power_spectrum(f_range, ap_params, gauss_params, nlv)
+freqs, spectrum = sim_power_spectrum(\
+    f_range, {'fixed' : ap_params}, {'gaussian' : gauss_params}, nlv)
 
 ###################################################################################################
 
@@ -222,7 +223,8 @@
 nlv = 0.025
 
 # Create a simulated power spectrum
-freqs, spectrum = sim_power_spectrum([1, 50], ap_params, gauss_params, nlv=nlv)
+freqs, spectrum = sim_power_spectrum(\
+    [1, 50], {'fixed' : ap_params}, {'gaussian' : gauss_params}, nlv=nlv)
 
 ###################################################################################################
 
@@ -276,8 +278,8 @@
 gauss_opts = param_sampler([[], [10, 0.5, 2], [10, 0.5, 2, 20, 0.3, 4]])
 
 # Simulate a group of power spectra
-freqs, power_spectra = sim_group_power_spectra(n_spectra, sim_freq_range,
-                                               ap_opts, gauss_opts, nlv)
+freqs, power_spectra = sim_group_power_spectra(\
+    n_spectra, sim_freq_range, {'fixed' : ap_opts}, {'gaussian' : gauss_opts}, nlv)
 
 ###################################################################################################
 
@@ -306,7 +308,7 @@
 ###################################################################################################
 
 # Find the index of the worst model fit from the group
-worst_fit_ind = np.argmax(fg.get_params('error'))
+worst_fit_ind = np.argmax(fg.get_params('metrics', 'error_mae'))
 
 # Extract this model fit from the group
 fm = fg.get_model(worst_fit_ind, regenerate=True)
@@ -335,11 +337,12 @@
 error_threshold = 0.010
 to_check = []
 for ind, res in enumerate(fg.results):
-    if res.error > error_threshold:
+    if res.metrics['error_mae'] > error_threshold:
         to_check.append(fg.get_model(ind, regenerate=True))
 
 # A more condensed version of the procedure above can be written like this:
-#to_check = [fg.get_model(ind, True) for ind, res in enumerate(fg) if res.error > error_threshold]
+#to_check = [fg.get_model(ind, True) for ind, res in enumerate(fg) \
+#                if res.metrics['error_mae'] > error_threshold]
 
 ###################################################################################################
 
diff --git a/tutorials/plot_09-FurtherAnalysis.py b/tutorials/plot_09-FurtherAnalysis.py
@@ -78,10 +78,10 @@
 set_random_seed(21)
 
 # Create some simulated power spectra
-freqs, spectra = sim_group_power_spectra(n_spectra=10,
-                                         freq_range=[3, 40],
-                                         aperiodic_params=param_sampler([[20, 2], [35, 1.5]]),
-                                         periodic_params=param_sampler([[], [10, 0.5, 2]]))
+freqs, spectra = sim_group_power_spectra(\
+    n_spectra=10, freq_range=[3, 40],
+    aperiodic_params={'fixed' : param_sampler([[20, 2], [35, 1.5]])},
+    periodic_params={'gaussian' : param_sampler([[], [10, 0.5, 2]])})
 
 ###################################################################################################
 
diff --git a/tutorials/plot_10-Reporting.py b/tutorials/plot_10-Reporting.py
@@ -105,7 +105,8 @@
 ###################################################################################################
 
 # Simulate an example power spectrum
-freqs, powers = sim_power_spectrum([1, 50], [0, 10, 1], [10, 0.25, 2], freq_res=0.25)
+freqs, powers = sim_power_spectrum(\
+    [1, 50], {'knee' : [0, 10, 1]}, {'gaussian' : [10, 0.25, 2]}, freq_res=0.25)
 
 # Initialize model object
 fm = SpectralModel(min_peak_height=0.1, peak_width_limits=[1, 6], aperiodic_mode='knee')
@@ -141,7 +142,8 @@
 ###################################################################################################
 
 # Simulate an example group of power spectra
-freqs, powers = sim_group_power_spectra(10, [1, 75], [0, 1], [10, 0.25, 2])
+freqs, powers = sim_group_power_spectra(\
+    10, [1, 75], {'fixed' : [0, 1]}, {'gaussian' : [10, 0.25, 2]})
 
 ###################################################################################################
 
