updated bgnetwork code including new plotting/analysis python modules…

…, param info for first paper (in params/), (pushed on matt's first day)

__init__.py

@@ -0,0 +1,6 @@
+#!usr/bin/env python
+import os
+import glob
+modules = glob.glob(os.path.dirname(__file__)+"/*.py")
+__all__ = [ os.path.basename(f)[:-3] for f in modules]
+_package_dir = os.path.dirname(os.path.realpath(__file__))

analyzefx.py

@@ -0,0 +1,173 @@
+import os
+import bgNetwork.netgen as ng
+import numpy as np
+import pandas as pd
+
+def print_results_summary(behavdf):
+    cond = behavdf.cond.iloc[0]
+    pLeft = behavdf.acc.mean()
+    corRT = behavdf[behavdf.acc==1].rtms.mean()
+    corMedRT = behavdf[behavdf.acc==1].rtms.median()
+    corRTvar = behavdf[behavdf.acc==1].rtms.sem()*2
+    errRT = 999.99; errRTvar = 0.0
+    if behavdf[behavdf.acc==0].shape[0]>1:
+        errRT = behavdf[behavdf.acc==0].rtms.mean()
+        errRTvar = behavdf[behavdf.acc==0].rtms.sem()*2
+    resultsSummary = 'Cond: {}\n\tMean CorRT:\t{:.2f} ({:.2f})\n\tMed CorRT:\t{:.2f}\n\tAvg ErrRT:\t{:.2f} ({:.2f})\n\tP(Left):\t{:.4f}'
+    print(resultsSummary.format(cond.upper(),corRT, corRTvar, corMedRT, errRT, errRTvar, pLeft))
+
+def print_trial_acc_rt(results, t=0):
+    bdata = ng.findOutputs(results[t])['decision made']
+    rt = bdata['time']
+    if bdata['pathvals'] is not None:
+        acc = ['Error' if bdata['pathvals'][0] else 'Correct'][0]
+        print('RT:\t{:.2f}ms\nAcc:\t{}\n'.format(rt,acc))
+        return 1
+    else:
+        return 0
+
+def get_msn_rates(sdf):
+    dL = sdf[sdf['population']=='dMSN0'].reset_index(drop=True)
+    dR = sdf[sdf['population']=='dMSN1'].reset_index(drop=True)
+    iL = sdf[sdf['population']=='iMSN0'].reset_index(drop=True)
+    iR = sdf[sdf['population']=='iMSN1'].reset_index(drop=True)
+    dlx = dL.fillna(method='pad', axis=1).convert_objects(convert_numeric=True)
+    drx = dR.fillna(method='pad', axis=1).convert_objects(convert_numeric=True)
+    ilx = iL.fillna(method='pad', axis=1).convert_objects(convert_numeric=True)
+    irx = iR.fillna(method='pad', axis=1).convert_objects(convert_numeric=True)
+    return [dlx, drx, ilx, irx]
+
+def get_avgMSN_traces(msnDFs, window=None):
+    rolltime = lambda df, w: df.rolling(w, axis=1).mean()
+    getMean = lambda df: df.groupby('cond2').mean().ix[:, 200:].values
+    getErr = lambda df: df.groupby('cond2').std().ix[:, 200:].values / 1.5
+    if window is not None:
+        msnDFs = [rolltime(msnPop, window) for msnPop in msnDFs]
+    ys = [getMean(msnPop) for msnPop in msnDFs]
+    ysErr = [getErr(msnPop) for msnPop in msnDFs]
+    return ys, ysErr
+
+def get_single_trial_ratedf(spikedf, trial=0):
+    spikedf = spikedf.drop(columns=['cond', 'cond2'])
+    sdf = spikedf[spikedf.trial==trial].T.reset_index(drop=True)
+    sdf.columns = sdf.iloc[0, :].values
+    sdf = sdf.drop(0)
+    sdf = sdf.fillna(np.nan)
+    return sdf
+
+def get_mean_firing_rates(spikedf, window=None):
+    avgdf = spikedf.groupby(['population']).mean().iloc[:, 1:].T.reset_index(drop=True)
+    return avgdf
+
+def get_firing_rates(results, window=None, getavg=False, cond='test'):
+    dflist = []
+    condABC = {'low': 'a', 'med': 'b', 'high': 'c', 'test':'x'}
+    for i, res in enumerate(results):
+        ratedf = res['popfreqs']
+        if window is not None:
+            ratedf = ratedf.rolling(window).mean()
+        ratedf = ratedf.T
+        if ratedf.shape[1]==0:
+            break
+        ratedf.insert(0, 'trial', i)
+        dflist.append(ratedf)
+    sdf = pd.concat(dflist).reset_index()
+    sdf = sdf.rename(columns={'index':'population'})
+    sdf = rename_populations(sdf)
+    sdf.insert(1, 'cond2', condABC[cond])
+    sdf.insert(1, 'cond', cond)
+    if getavg:
+        return get_mean_firing_rates(sdf)
+    return sdf
+
+def analyze_network_behavior(results, preset, stimArray, cond='test', outdir=None):
+    rtms, choice, acc, decisions = [], [], [], []
+    condABC = {'low': 'a', 'med': 'b', 'high': 'c'}
+    condLevel = {'low': 1, 'med': 2, 'high': 3, 'test': 0}
+    for result in results:
+        decisions.append(ng.findOutputs(result)['decision made'])
+    for decision in decisions:
+        if decision['pathvals'] is None:
+            choice.append(np.nan)
+            acc.append(np.nan)
+            rtms.append(np.nan)
+            continue
+        elif decision['pathvals'] == [0]:
+            choice.append(0)
+            acc.append(1)
+        else:
+            choice.append(1)
+            acc.append(0)
+        rtms.append(decision['delay'])
+    behav_arr = [np.asarray(arr) for arr in [choice, acc, rtms]]
+    behavdf = pd.DataFrame(dict(zip(['choice', 'acc', 'rtms'], behav_arr)))
+    behavdf['rtms'] = behavdf.rtms.values + 200
+    behavdf['rt'] = behavdf.rtms.values * .001
+    behavdf.insert(0, 'trial', np.arange(1, behavdf.shape[0]+1))
+    behavdf.insert(0, 'response', behavdf.acc.values)
+    behavdf.insert(0, 'cond', cond)
+    behavdf.insert(0, 'stim', stimArray[:behavdf.shape[0]])
+    behavdf.insert(0, 'subj_idx', 1)
+    behavdf['cond2'] = condABC[cond]
+    behavdf['level'] = condLevel[cond]
+    dL,dR = preset[0]['cxd']['mult']
+    iL,iR = preset[0]['cxi']['mult']
+    behavdf['wdL'] = dL
+    behavdf['wdR'] = dR
+    behavdf['wiL'] = iL
+    behavdf['wiR'] = iR
+    behavdf = behavdf[['subj_idx', 'cond', 'cond2', 'level', 'trial', 'stim', 'wdL', 'wiL', 'wdR', 'wiR', 'response', 'acc', 'choice', 'rtms', 'rt']]
+    print_results_summary(behavdf)
+    if outdir is not None:
+        behavdfName = os.path.join(outdir,'cbgt_behavdf_{}.csv'.format(cond))
+        outdir = os.path.join(outdir, behavdfName)
+        behavdf.to_csv(behavdfName, index=False)
+    return behavdf
+
+
+def make_wts_df(presets=None, conds=['low', 'med', 'high'], ntrials=1000):
+
+    if presets is None:
+        presetLow = [{'cxd': {'dest': [0, 1], 'mult': [1.02, 0.99], 'src': [0, 1]},
+              'cxi': {'dest': [0, 1], 'mult': [1.0, 1.0], 'src': [0, 1]}}]
+        presetMed = [{'cxd': {'dest': [0, 1], 'mult': [1.035, 0.98], 'src': [0, 1]},
+                      'cxi': {'dest': [0, 1], 'mult': [1.0, 1.0], 'src': [0, 1]}}]
+        presetHi = [{'cxd': {'dest': [0, 1], 'mult': [1.045, 0.97], 'src': [0, 1]},
+                     'cxi': {'dest': [0, 1], 'mult': [1.0, 1.0], 'src': [0, 1]}}]
+        presets = [presetLow, presetMed, presetHi]
+
+    condlist, dleft, dright, ileft, iright = [], [], [], [], []
+    for i, pre in enumerate(presets):
+        dL,dR = pre[0]['cxd']['mult']
+        iL,iR = pre[0]['cxi']['mult']
+        condlist.extend([conds[i]]*ntrials)
+        dleft.extend([dL]*ntrials)
+        dright.extend([dR]*ntrials)
+        ileft.extend([iL]*ntrials)
+        iright.extend([iR]*ntrials)
+    weightdf = pd.DataFrame({'cond':condlist, 'wdL': dleft, 'wiL':ileft, 'wdR':dright, 'wiR':iright})[['cond', 'wdL', 'wiL', 'wdR', 'wiR']]
+    return weightdf
+
+def rename_populations(sdf):
+    popNames = []
+    oldPopNames = sdf.population.unique().tolist()
+    for pop in oldPopNames:
+        poplist = pop.split('_')
+        if len(poplist)==2:
+            popName = poplist[0]
+        else:
+            popName = ''.join(poplist[:2])
+        if 'D1STR' in popName:
+            popName = 'dMSN'+popName[-1]
+        elif 'D2STR' in popName:
+            popName = 'iMSN'+popName[-1]
+        elif 'STNE' in popName:
+            popName = 'STN'+popName[-1]
+        elif 'GPeP' in popName:
+            popName = 'GPe'+popName[-1]
+        elif 'Time (ms)' in popName:
+            popName = 'time'
+        popNames.append(popName)
+
+    sdfNew = sdf.replace({'population': dict(zip(oldPopNames, popNames))})
+    return sdfNew