Merge pull request #1 from jkosciessa/dev-kosciessa

Minor updates to the doc pages

.gitignore

@@ -26,6 +26,7 @@ share/python-wheels/
 .installed.cfg
 *.egg
 MANIFEST
+*.DS_STORE
 
 # PyInstaller
 #  Usually these files are written by a python script from a template
@@ -119,6 +120,9 @@ venv.bak/
 
 # mkdocs documentation
 /site
+/doc/_build/*
+/doc/auto_examples/*
+/doc/generated/*
 
 # mypy
 .mypy_cache/
@@ -127,3 +131,6 @@ dmypy.json
 
 # Pyre type checker
 .pyre/
+
+# examples
+/examples/*_data/*

doc/API/datatypes.rst

@@ -1,12 +1,12 @@
 Data Types
 ===================================
 
-.. autoclass:: whobpyt.datatypes.Recording
+.. autoclass:: whobpyt.datatypes.Timeseries
    :members:
    :undoc-members:
    :special-members: __init__
 
-.. autoclass:: whobpyt.datatypes.par
+.. autoclass:: whobpyt.datatypes.Parameter
    :members:
    :undoc-members:
    :special-members: __init__

doc/API/models.rst

@@ -1,27 +1,12 @@
 Models
 ===================================
 
-.. autoclass:: whobpyt.models.JansenRit.RNNJANSEN
-   :members: info, createIC, createDelayIC, setModelParameters, forward
+.. autoclass:: whobpyt.models.jansen_rit.JansenRitParams
+   :members: info, createIC, createDelayIC, setModelSCParameters, forward
    :undoc-members:
    :special-members: __init__
 
-.. autoclass:: whobpyt.models.RWW.RNNRWW
-   :members: info, createIC, setModelParameters, forward
-   :undoc-members:
-   :special-members: __init__
-
-.. autoclass:: whobpyt.models.RWWEI2.RWWEI2
-   :members: info, createIC, setModelParameters, forward
-   :undoc-members:
-   :special-members: __init__
-
-.. autoclass:: whobpyt.models.BOLD.BOLD_Layer
-   :members: info, createIC, setModelParameters, forward
-   :undoc-members:
-   :special-members: __init__
-
-.. autoclass:: whobpyt.models.EEG.EEG_Layer
-   :members: info, createIC, setModelParameters, forward
+.. autoclass:: whobpyt.models.wong_wang.ReducedWongWangModel
+   :members: info, createIC, createDelayIC, setModelSCParameters, forward
    :undoc-members:
    :special-members: __init__

doc/Makefile

@@ -9,7 +9,6 @@
 # You can set these variables from the command line.
 SPHINXOPTS    =
 SPHINXBUILD   = sphinx-build
-#SPHINXPROJ    = fooof
 SPHINXPROJ    = whobpyt
 SOURCEDIR     = .
 BUILDDIR      = _build
@@ -50,7 +49,6 @@ install:
 	#   --no-checkout just fetches the root folder without content
 	#   --depth 1 is a speed optimization since we don't need the history prior to the last commit
 	#   -b gh-pages fetches only the branch for the gh-pages
-	#git clone -b gh-pages --single-branch --no-checkout --depth 1 https://github.com/fooof-tools/fooof _build/gh_pages
 	git clone -b gh-pages --single-branch --no-checkout --depth 1 [email protected]:griffithslab/whobpyt _build/gh_pages
 	# A .nojekyll file tells Github pages to bypass Jekyll processing
 	touch _build/gh_pages/.nojekyll

examples/eg__RWW_example.py

@@ -0,0 +1,177 @@
+"""
+
+==================================
+Fitting Wong_Wang model (HCP data)
+==================================
+
+"""
+
+# sphinx_gallery_thumbnail_number = 1
+
+# %%
+# Importage
+# ---------
+#
+
+# os stuff
+import os
+import sys
+sys.path.append('..')
+
+# whobpyt stuff
+import whobpyt
+from whobpyt.datatypes import Parameter, Timeseries
+from whobpyt.models.wong_wang import ReducedWongWangModel, ReducedWongWangParams
+from whobpyt.optimization.custom_cost_RWW import CostsRWW
+from whobpyt.run import ModelFitting
+from whobpyt.datasets import fetch_hcpl2k8
+
+# array and pd stuff
+import numpy as np
+import pandas as pd
+
+# viz stuff
+import matplotlib.pyplot as plt
+
+# gdown
+import gdown
+
+# %%
+# define destination path and download the data
+
+
+"""
+des_dir = '../'
+if not os.path.exists(des_dir):
+    os.makedirs(des_dir)  # create folder if it does not exist
+url = "https://drive.google.com/drive/folders/18smy3ElTd4VksoL4Z15dhwT5l3yjk6xS"
+os.chdir(des_dir)
+gdown.download_folder(url, remaining_ok = True)
+"""
+
+# Go to examples folder
+#os.chdir('examples')
+
+
+# Download the data for this example to default location ~/.whobpyt/data
+data_dir = fetch_hcpl2k8()
+
+base_dir = os.path.join(data_dir, 'HCP')
+
+# %%
+#get subject list
+#base_dir = '../HCP/'
+
+#subs =sorted([sc_file[-10:-4] for sc_file in os.listdir(base_dir) if sc_file[:8] == 'weights_'])
+
+sub = '100307'
+
+# %%
+# define options for wong-wang model
+node_size = 83
+mask = np.tril_indices(node_size, -1)
+num_epochs = 5
+TPperWindow = 20
+step_size = 0.05
+input_size = 2
+tr = 0.75
+repeat_size = 5
+
+# %%
+# load raw data and get SC empirical BOLD and FC
+sc_file = os.path.join(base_dir, 'weights_' + sub + '.txt')
+ts_file = os.path.join(base_dir, sub + '_rfMRI_REST1_LR_hpc200_clean__l2k8_sc33_ts.pkl')  # out_dir+'sub_'+sub+'simBOLD_idt.txt'
+
+sc = np.loadtxt(sc_file)
+SC = (sc + sc.T) * 0.5
+sc = np.log1p(SC) / np.linalg.norm(np.log1p(SC))
+
+ts_pd = pd.read_pickle(ts_file)
+ts = ts_pd.values
+ts = ts / np.max(ts)
+fc_emp = np.corrcoef(ts.T)
+
+# %%
+# prepare data structure for model
+print(ts.T.shape)
+fMRIstep = tr
+data_mean = Timeseries(ts.T, fMRIstep)
+
+# %%
+# get model parameters structure and define the fitted parameters by setting non-zero variance for the model
+params = ReducedWongWangParams(g=Parameter(400, 400, 1/np.sqrt(10), True), g_EE=Parameter(1.5, 1.5, 1/np.sqrt(50), True), \
+g_EI =Parameter(0.8, 0.8, 1/np.sqrt(50), True), \
+                   g_IE=Parameter(0.6, 0.6, 1/np.sqrt(50), True), I_0 =Parameter(0.2), \
+                   std_in=Parameter(np.log(0.1), np.log(0.1), 0.1, True, True, lb =0.02))
+
+# %%
+# call model want to fit
+model = ReducedWongWangModel(params, node_size =node_size, TRs_per_window =TPperWindow, step_size=step_size, tr=tr, sc=sc, use_fit_gains=True)
+
+# %%
+# create objective function
+ObjFun = CostsRWW(model)
+
+# %%
+# call model fit
+F = ModelFitting(model, ObjFun)
+
+
+# %%
+# Model Training
+# ---------------------------------------------------
+#
+F.train(u = 0, empRecs = [data_mean], num_epochs = num_epochs, TPperWindow = TPperWindow, learningrate = 0.05)
+
+# %%
+# Plots of loss over Training
+plt.plot(np.arange(1,len(F.trainingStats.loss)+1), F.trainingStats.loss)
+plt.title("Total Loss over Training Epochs")
+
+# %%
+# Plots of parameters values over Training
+plt.plot(F.trainingStats.fit_params['g_EE'], label = "g_EE")
+plt.plot(F.trainingStats.fit_params['g_EI'], label = "g_EI")
+plt.plot(F.trainingStats.fit_params['g_IE'], label = "g_IE")
+plt.legend()
+plt.title("Local Coupling Variables Changing Over Training Epochs")
+
+# %%
+# Model Evaluation (with 20 window for warmup)
+# ---------------------------------------------------
+#
+F.evaluate(u = 0, empRec = data_mean, TPperWindow = TPperWindow, base_window_num = 20)
+
+# %%
+# Plot SC and fitted SC
+
+fig, ax = plt.subplots(1, 2, figsize=(5, 4))
+im0 = ax[0].imshow(sc, cmap='bwr', vmin = 0.0, vmax = 0.05)
+ax[0].set_title('The empirical SC')
+fig.colorbar(im0, ax=ax[0], fraction=0.046, pad=0.04)
+im1 = ax[1].imshow(F.model.sc_fitted.detach().numpy(), cmap='bwr', vmin = 0.0, vmax = 0.05)
+ax[1].set_title('The fitted SC')
+fig.colorbar(im1, ax=ax[1], fraction=0.046, pad=0.04)
+plt.show()
+
+# %%
+# Plot E I and simulated BOLD
+fig, ax = plt.subplots(1, 3, figsize=(12, 8))
+ax[0].plot(F.lastRec['E'].npTS().T)
+ax[0].set_title('Test: E')
+ax[1].plot(F.lastRec['I'].npTS().T)
+ax[1].set_title('Test: I')
+ax[2].plot(F.lastRec['bold'].npTS().T)
+ax[2].set_title('Test: BOLD')
+plt.show()
+
+# %%
+# Plot the FC and the test FC
+fig, ax = plt.subplots(1, 2, figsize=(5, 4))
+im0 = ax[0].imshow(fc_emp, cmap='bwr')
+ax[0].set_title('The empirical FC')
+fig.colorbar(im0, ax=ax[0], fraction=0.046, pad=0.04)
+im1 = ax[1].imshow(np.corrcoef(F.lastRec['bold'].npTS()), cmap='bwr')
+ax[1].set_title('The simulated FC')
+fig.colorbar(im1, ax=ax[1], fraction=0.046, pad=0.04)
+plt.show()

examples/eg__replicate_Momi2023.py

@@ -113,7 +113,7 @@
 output_size = eeg_data.shape[0]
 batch_size = 20
 step_size = 0.0001
-num_epochs = 20 #2 # num_epochs = 20
+num_epochs = 2 # num_epochs = 20
 tr = 0.001
 state_size = 6
 base_batch_num = 20
@@ -131,35 +131,35 @@
 # get model parameters structure and define the fitted parameters by setting non-zero variance for the model
 lm = np.zeros((output_size,200))
 lm_v = np.zeros((output_size,200))
-params = JansenRitParams(A = par(3.25), 
-                         a= par(100,100, 2, True), 
-                         B = par(22), 
+params = JansenRitParams(A=par(3.25), 
+                         a=par(100,100, 2, True), 
+                         B=par(22), 
                          b = par(50, 50, 1, True),
-                         g=par(500,500,2, True), 
-                         g_f=par(10,10,1, True), 
-                         g_b=par(10,10,1, True),
+                         g = par(500,500,2, True), 
+                         g_f = par(10,10,1, True), 
+                         g_b = par(10,10,1, True),
                          c1 = par(135, 135, 1, True), 
                          c2 = par(135*0.8, 135*0.8, 1, True), 
                          c3 = par(135*0.25, 135*0.25, 1, True),
                          c4 = par(135*0.25, 135*0.25, 1, True), 
-                         std_in= par(np.log(10), np.log(10), .1, True, True), 
-                         vmax= par(5), 
-                         v0=par(6), 
-                         r=par(0.56),
-                         y0=par(-2, -2, 1/4, True),
+                         std_in = par(np.log(10), np.log(10), .1, True, True), 
+                         vmax = par(5), 
+                         v0 = par(6), 
+                         r = par(0.56),
+                         y0 = par(-2, -2, 1/4, True),
                          mu = par(np.log(1.5),
                                   np.log(1.5), .1, True, True, lb=0.1), 
-                         k =par(5., 5., 0.2, True, lb=1), 
-                         k0=par(0),
+                         k = par(5., 5., 0.2, True, lb=1), 
+                         k0 = par(0),
                          cy0 = par(50, 50, 1, True), 
-                         ki=par(ki0), 
-                         lm=par(lm, lm, 1 * np.ones((output_size, node_size))+lm_v, True)
+                         ki = par(ki0), 
+                         lm = par(lm, lm, 1 * np.ones((output_size, node_size))+lm_v, True)
                          )
 
 
 # %%
 # call model want to fit
-model = JansenRitModel(params, 
+model = JansenRitModel(params,
                        node_size=node_size, 
                        TRs_per_window=batch_size, 
                        step_size=step_size, 
@@ -169,7 +169,7 @@
                        lm=lm, 
                        dist=dist, 
                        use_fit_gains=True, 
-                       use_fit_lfm = False)
+                       use_fit_lfm=False)
 
 
 
@@ -195,7 +195,7 @@
 # %%
 # load in a previously completed model fitting results object
 full_run_fname = os.path.join(data_dir, 'Subject_1_low_voltage_fittingresults_stim_exp.pkl')
-F = pickle.load(open(full_run_fname, 'rb'))
+#F = pickle.load(open(full_run_fname, 'rb'))
 F.evaluate(u = u, empRec = data_mean, TPperWindow = batch_size, base_window_num = 20)
 
 

requirements.txt

@@ -14,7 +14,7 @@ mne
 # Docs requirements
 sphinx
 #sphinx==3.1.1
-sphinx-gallery==0.8.1
+sphinx_gallery==0.8.1
 sphinx_rtd_theme #==0.5.0
 sphinx-tabs==1.3.0
 sphinx-copybutton==0.3.1

setup.py

@@ -29,6 +29,6 @@
         "License :: OSI Approved :: MIT License",
         "Operating System :: OS Independent",
     ],
-    python_requires='>=3.9',
+    python_requires='>=3.8',
 )
 

whobpyt/models/jansen_rit/jansen_rit.py

@@ -170,7 +170,7 @@ class JansenRitModel(AbstractNeuralModel):
     createDelayIC(self, ver):
         Creates the initial conditions for the delays.
 
-    setModelParameters(self):    
+    setModelSCParameters(self):    
         Sets the parameters of the model.
     
     forward(input, noise_out, hx)
@@ -244,7 +244,7 @@ def __init__(self,
         self.params = params
         self.output_size = lm.shape[0]  # number of EEG channels
 
-        self.setModelParameters()
+        self.setModelParameters() # defined in datatypes/abstract_neural_model.py
         self.setModelSCParameters()
 
 

whobpyt/models/wong_wang/__init__.py

@@ -0,0 +1 @@
+from .wong_wang import ReducedWongWangModel, ReducedWongWangParams