build: added Windows OS support and compatibility

- Added Windows-specific installation instructions
- Fix path separators to use OS-dependent ones
- Updated dependency packages for simulation examples

Author: laurator

README.md

@@ -35,7 +35,7 @@ parameter inference techniques based on [sbi](https://sbi-dev.github.io/sbi)) wi
 inverse modelling pipeline. Designed for modularity, it adapts to your needs without constraints.
 
 
-# Installation
+# Installation on Linux
 
 `ncpi` requires Python 3.10 or higher. To install `ncpi`, you can use pip. The package is available on PyPI, so you can 
 install it directly from there.
@@ -73,6 +73,31 @@ pip uninstall scikit-learn numpy -y
 pip install scikit-learn==1.5.0 numpy
 ```
 
+# Installation on Windows
+
+To be able to install all dependencies of `ncpi` in Windows, first you have to install [Windows Subsystem for Linux (WSL)](https://documentation.ubuntu.com/wsl/stable/howto/install-ubuntu-wsl2/).
+
+After the WSL installation, we strongly recommend to install the latest updates by running the following commands within the Ubuntu terminal:
+
+```bash
+$ sudo apt update
+$ sudo apt upgrade -y
+```
+
+Once Ubuntu is up and running in WSL, [conda can be installed there](https://docs.conda.io/projects/conda/en/stable/user-guide/install/linux.html). Now you can follow the rest of the instructions of the [installation of `ncpi` in Linux](#Installation-on-Linux). 
+
+If you encounter the error *Failed building wheel for pycatch22* when installing `pip install ncpi`, install:
+
+```bash
+$ conda install -c conda-forge pycatch22
+```
+
+If the error still persists, install the following dependencies:
+
+```bash
+$ sudo apt install -y build-essential python3-dev
+```
+
 # Folder Structure
 
 - `ncpi/`: Contains the source code for the library, organized into modules and classes.

examples/EEG_AD/EEG_AD.py

@@ -11,9 +11,9 @@
 # Select either raw EEG data or source-reconstructed EEG data. This study used the raw EEG data for all analyses.
 raw = True
 if raw:
-    data_path = '/DATA/empirical_datasets/POCTEP_data/CLEAN/SENSORS'
+    data_path = os.path.join(os.sep, 'DATA', 'empirical_datasets', 'POCTEP_data', 'CLEAN', 'SENSORS')
 else:
-    data_path = '/DATA/empirical_datasets/POCTEP_data/CLEAN/SOURCES/dSPM/DK'
+    data_path = os.path.join(os.sep, 'DATA', 'empirical_datasets', 'POCTEP_data', 'CLEAN', 'SOURCES', 'dSPM', 'DK')
 
 # Choose to either download data from Zenodo (True) or load it from a local path (False).
 # Important: the zenodo downloads will take a while, so if you have already downloaded the data, set this to False and
@@ -208,12 +208,12 @@ def create_POCTEP_dataframe(data_path):
             print(f'--- Sensor: {sensor}')
 
             shutil.copy(
-                os.path.join(zenodo_dir_sim, 'ML_models/EEG', sensor, method, 'model'),
+                os.path.join(zenodo_dir_sim, 'ML_models', 'EEG', sensor, method, 'model'),
                 os.path.join('data', 'model.pkl')
             )
 
             shutil.copy(
-                os.path.join(zenodo_dir_sim, 'ML_models/EEG', sensor, method, 'scaler'),
+                os.path.join(zenodo_dir_sim, 'ML_models', 'EEG', sensor, method, 'scaler'),
                 os.path.join('data', 'scaler.pkl')
             )
 
@@ -237,5 +237,4 @@ def create_POCTEP_dataframe(data_path):
                                                    np.array(others.tolist())), axis=1)).tolist()
 
         # Save the data including predictions of (E/I)_net
-        emp_data.to_pickle(os.path.join('data', method, 'emp_data_reduced.pkl'))
-            
+        emp_data.to_pickle(os.path.join('data', method, 'emp_data_reduced.pkl'))

examples/EEG_AD/figures/EEG_predictions.py

@@ -6,7 +6,7 @@
 import ncpi
 
 # Set the path to the results folder
-results_path = '../data'
+results_path = os.path.join('..', 'data')
 
 # Select the statistical analysis method ('cohen', 'lmer')
 statistical_analysis = 'lmer'
@@ -82,7 +82,7 @@ def append_lmer_results(lmer_results, group, elec, p_value_th, data_lmer):
         if row == 2 or row == 3:
             method = 'power_spectrum_parameterization_1'
         try:
-            data = pd.read_pickle(os.path.join('../data', method, 'emp_data_reduced.pkl'))
+            data = pd.read_pickle(os.path.join(results_path, method, 'emp_data_reduced.pkl'))
 
         except Exception as e:
             print(f'Error loading data for {method}: {e}')
@@ -238,6 +238,6 @@ def append_lmer_results(lmer_results, group, elec, p_value_th, data_lmer):
     fig1.add_artist(tauexc_line_f)
     fig1.add_artist(tauinh_line_f)
 
-    fig1.savefig('EEG_predictions.png')
+    fig1.savefig(f'EEG_predictions_{statistical_analysis}.png')
 
 

examples/LFP_developing_brain/LFP_developing_brain.py

@@ -22,8 +22,8 @@
 zenodo_URL_emp = "https://zenodo.org/api/records/15382047"
 
 # Paths to zenodo files
-zenodo_dir_sim = "zenodo_sim_files"
-zenodo_dir_emp= "zenodo_emp_files"
+zenodo_dir_sim = os.path.join("zenodo_sim_files")
+zenodo_dir_emp= os.path.join("zenodo_emp_files")
 
 # Methods used to compute the features
 all_methods = ['catch22','power_spectrum_parameterization_1']
@@ -79,18 +79,18 @@ def load_inference_data(method, X, theta, zenodo_dir_sim, ML_model):
         folder = 'SBI'
 
     shutil.copy(
-        os.path.join(zenodo_dir_sim, 'ML_models/4_param', folder, method, 'scaler'),
+        os.path.join(zenodo_dir_sim, 'ML_models', '4_param', folder, method, 'scaler'),
         os.path.join('data', 'scaler.pkl')
     )
 
     shutil.copy(
-        os.path.join(zenodo_dir_sim, 'ML_models/4_param', folder, method, 'model'),
+        os.path.join(zenodo_dir_sim, 'ML_models', '4_param', folder, method, 'model'),
         os.path.join('data', 'model.pkl')
     )
 
     if ML_model == 'NPE':
         shutil.copy(
-            os.path.join(zenodo_dir_sim, 'ML_models/4_param', folder, method,
+            os.path.join(zenodo_dir_sim, 'ML_models', '4_param', folder, method,
                             'density_estimator'),
             os.path.join('data', 'density_estimator.pkl')
         )
@@ -101,13 +101,13 @@ def load_inference_data(method, X, theta, zenodo_dir_sim, ML_model):
 @timer("Loading LFP data.")
 def load_empirical_data(zenodo_dir_emp):
     # Load empirical data
-    file_list = os.listdir(os.path.join(zenodo_dir_emp, 'development_EI_decorrelation/baseline/LFP'))
+    file_list = os.listdir(os.path.join(zenodo_dir_emp, 'development_EI_decorrelation', 'baseline', 'LFP'))
     emp_data = {'LFP': [], 'fs': [], 'age': []}
 
     for i,file_name in enumerate(file_list):
         print(f'\r Progress: {i+1} of {len(file_list)} files loaded', end='', flush=True)
         structure = scipy.io.loadmat(os.path.join(os.path.join(zenodo_dir_emp,
-                                                                'development_EI_decorrelation/baseline/LFP'),
+                                                                'development_EI_decorrelation', 'baseline', 'LFP'),
                                                     file_name))
         LFP = structure['LFP']['LFP'][0,0]
         sum_LFP = np.sum(LFP, axis=0)  # sum LFP across channels

examples/LFP_developing_brain/figures/LFP_predictions.py

@@ -7,17 +7,17 @@
 import ncpi
 
 # Folder with parameters of LIF model simulations
-sys.path.append(os.path.join(os.path.dirname(__file__), '../../simulation/Hagen_model/simulation/params'))
+sys.path.append(os.path.join(os.path.dirname(__file__), '..', '..', 'simulation', 'Hagen_model', 'simulation', 'params'))
 
 # Path to the folder with prediction results
-pred_results = '../data'
+pred_results = os.path.join('..', 'data')
 
 # Calculate new firing rates (True) or load them from file if they already exist (False). If firing rates do not
 # exist, they will not be plotted.
 compute_firing_rate = False
 
 # Path to saved firing rates
-fr_path = './data'
+fr_path = os.path.join('.', 'data')
 
 # Number of samples to draw from the predictions for computing the firing rates
 n_samples = 50
@@ -118,23 +118,23 @@
                 LIF_params['J_ext'] = J_ext
 
                 # Create a Simulation object
-                sim = ncpi.Simulation(param_folder='../../simulation/Hagen_model/simulation/params',
-                                      python_folder='../../simulation/Hagen_model/simulation/python',
-                                      output_folder='../../simulation/Hagen_model/simulation/output')
+                sim = ncpi.Simulation(param_folder = os.path.join('../../simulation/Hagen_model/simulation/params'),
+                                      python_folder = os.path.join('../../simulation/Hagen_model/simulation/python'),
+                                      output_folder = os.path.join('../../simulation/Hagen_model/simulation/output'))
 
                 # Save parameters to a pickle file
-                with open(os.path.join('../../simulation/Hagen_model/simulation/output', 'network.pkl'), 'wb') as f:
+                with open(os.path.join('..', '..', 'simulation', 'Hagen_model', 'simulation', 'output', 'network.pkl'), 'wb') as f:
                     pickle.dump(LIF_params, f)
 
                 # Run the simulation
                 sim.simulate('simulation.py', 'simulation_params.py')
 
                 # Load spike times
-                with open(os.path.join('../../simulation/Hagen_model/simulation/output', 'times.pkl'), 'rb') as f:
+                with open(os.path.join('..', '..', 'simulation', 'Hagen_model', 'simulation', 'output', 'times.pkl'), 'rb') as f:
                     times = pickle.load(f)
 
                 # Load tstop
-                with open(os.path.join('../../simulation/Hagen_model/simulation/output', 'tstop.pkl'), 'rb') as f:
+                with open(os.path.join('..', '..', 'simulation', 'Hagen_model', 'simulation', 'output', 'tstop.pkl'), 'rb') as f:
                     tstop = pickle.load(f)
 
                 # Transient period
@@ -155,15 +155,15 @@
 if compute_firing_rate:
     if not os.path.exists('data'):
         os.makedirs('data')
-    with open('data/firing_rates_preds.pkl', 'wb') as f:
+    with open('data', 'firing_rates_preds.pkl', 'wb') as f:
         pickle.dump(firing_rates, f)
-    with open('data/IDs.pkl', 'wb') as f:
+    with open('data', 'IDs.pkl', 'wb') as f:
         pickle.dump(IDs, f)
 else:
     try:
-        with open(os.path.join(fr_path,'firing_rates_preds.pkl'), 'rb') as f:
+        with open(os.path.join(fr_path, 'firing_rates_preds.pkl'), 'rb') as f:
             firing_rates = pickle.load(f)
-        with open(os.path.join(fr_path,'IDs.pkl'), 'rb') as f:
+        with open(os.path.join(fr_path, 'IDs.pkl'), 'rb') as f:
             IDs = pickle.load(f)
     except FileNotFoundError:
         print('Firing rates not found.')
@@ -383,5 +383,5 @@
 ax.text(0.5, 0.49, y_labels[1], color = 'blue', alpha = 0.5, fontsize = 10, ha='center')
 
 # Save the figure
-plt.savefig('LFP_predictions.png', bbox_inches='tight')
+plt.savefig(f'LFP_predictions_{statistical_analysis}.png', bbox_inches='tight')
 # plt.show()

examples/LFP_developing_brain/figures/emp_features.py

@@ -3,7 +3,7 @@
 import matplotlib.pyplot as plt
 
 # Path to the folder with prediction results
-pred_results = '../data'
+pred_results = os.path.join('..', 'data')
 
 # Names of catch22 features
 try:

examples/simulation/Hagen_model/figures/SBI_results.py

@@ -17,7 +17,7 @@
 compute_metrics = True
 
 # Path to the local directory where the metrics and posteriors will be saved
-result_folder = 'SBI_results'
+result_folder = os.path.join('SBI_results')
 
 # Choose whether to use a held-out dataset or folds from RepeatedKFold
 use_held_out_data = True
@@ -34,7 +34,7 @@
 zenodo_URL_sim = "https://zenodo.org/api/records/15351118"
 
 # Paths to zenodo files
-zenodo_dir_sim = "/DATA/zenodo_sim_files"
+zenodo_dir_sim = os.path.join(os.sep, 'DATA', 'zenodo_sim_files')
 
 # Download simulation data and ML models
 if zenodo_dw_sim:
@@ -105,10 +105,10 @@ def create_white_to_color_cmap(color):
 
 # Path to ML models trained based on a held-out dataset approach
 if use_held_out_data:
-    ML_path = os.path.join(zenodo_dir_sim, 'ML_models/held_out_data_models')
+    ML_path = os.path.join(zenodo_dir_sim, 'ML_models', 'held_out_data_models')
 # Path to ML models trained based on a RepeatedKFold approach
 else:
-    ML_path = os.path.join(zenodo_dir_sim, 'ML_models/4_param')
+    ML_path = os.path.join(zenodo_dir_sim, 'ML_models', '4_param')
 
 # Limits of histograms
 lims = [[-15, 15], [-2, 5], [-2, 12], [0, 60]]

examples/simulation/Hagen_model/figures/example_full_pipeline.py

@@ -10,7 +10,7 @@
 from ncpi import tools
 
 # Path to parameters of the LIF network model
-sys.path.append(os.path.join(os.path.dirname(__file__), '../simulation/params'))
+sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'simulation', 'params'))
 
 # Choose to either download files and precomputed outputs used in simulations of the reference multicompartment neuron
 # network model (True) or load them from a local path (False)
@@ -20,7 +20,7 @@
 zenodo_URL_mult = "https://zenodo.org/api/records/15429373"
 
 # Zenodo directory where the data is stored (must be an absolute path to correctly load morphologies in NEURON)
-zenodo_dir = '/DATA/multicompartment_neuron_network'
+zenodo_dir = os.path.join(os.sep, 'DATA', 'multicompartment_neuron_network')
 
 # Set to True to run new simulations of the LIF network model, or False to load precomputed results from a pickle file
 # located in a 'data' folder.

examples/simulation/Hagen_model/figures/sim_features.py

@@ -17,7 +17,7 @@
 zenodo_URL_sim = "https://zenodo.org/api/records/15351118"
 
 # Paths to zenodo files
-zenodo_dir_sim = "zenodo_sim_files"
+zenodo_dir_sim = os.path.join("zenodo_sim_files")
 
 # Download simulation data and ML models
 if zenodo_dw_sim:

examples/simulation/Hagen_model/figures/sim_predictions.py

@@ -16,7 +16,7 @@
 zenodo_URL_sim = "https://zenodo.org/api/records/15351118"
 
 # Paths to zenodo files
-zenodo_dir_sim = "zenodo_sim_files"
+zenodo_dir_sim = os.path.join("zenodo_sim_files")
 
 # ML model used to compute the predictions (MLPRegressor or Ridge)
 ML_model = 'MLPRegressor'
@@ -69,10 +69,10 @@ def hellinger_distance(p, q):
 all_theta = {}
 for method in all_methods:
     try:
-        with open(os.path.join(zenodo_dir_sim,'ML_models/held_out_data_models', folder, method,
+        with open(os.path.join(zenodo_dir_sim,'ML_models', 'held_out_data_models', folder, method,
                                'predictions'), 'rb') as file:
             all_preds[method] = np.array(pickle.load(file))
-        with open(os.path.join(zenodo_dir_sim,'ML_models/held_out_data_models', 'datasets', method,
+        with open(os.path.join(zenodo_dir_sim,'ML_models', 'held_out_data_models', 'datasets', method,
                                'held_out_dataset'), 'rb') as file:
             X_test, theta_test = pickle.load(file)
             all_theta[method] = np.array(theta_test)