# plt.scatter(cms, av, alpha=0.1)
# plt.ylabel('Angle Velocity')
# plt.xlabel('CMS Velocity')
# plt.subplot(212)
# plt.plot(cms, label='CMS velocity')
# plt.plot(av, label='Angle Velocity', alpha=0.7)
# plt.legend()
# plt.show()
### play with LASSO
beh = 'Eigenworm3'
data = dh.loadData(folder, dataPars)
results = dr.runLasso(data, pars, testIndices, trainingsIndices, plot = 1, behaviors = [beh])#, 'Eigenworm3'])
results = dr.runElasticNet(data, pars, testIndices, trainingsIndices, plot = 1, behaviors = [beh])#, 'Eigenworm3'])
###############################################
#
# Test how stable LASSO is in response to median filtering
#
##############################################
medians = np.arange(1,15,2)
medianFiltData = np.zeros((len(medians),3))
beh = 'Eigenworm3'
dataPars['savGolayWindow'] = 13
for ind, medfilt in enumerate(medians):
dataPars['medianWindow'] = medfilt
data = dh.loadData(folder, dataPars)
results = dr.runLasso(data, pars, testIndices, trainingsIndices, plot = 1, behaviors = [beh])#, 'Eigenworm3'])
#mp.plotPCAresults3D(dataSets, resultDict, keyList, pars, col = 'etho', flag = 'LASSO')
plt.show()
#%%
###############################################
#
# linear regression using elastic Net
#
##############################################
if elasticnet:
for kindex, key in enumerate(keyList):
print 'Running Elastic Net', key
splits = resultDict[key]['Training']
resultDict[key]['ElasticNet'] = dr.runElasticNet(dataSets[key],
pars,
splits,
plot=1,
behaviors=behaviors)
# calculate how much more neurons contribute
tmpDict = dr.scoreModelProgression(
dataSets[key],
resultDict[key],
splits,
pars,
fitmethod='ElasticNet',
behaviors=behaviors,
)
for tmpKey in tmpDict.keys():
resultDict[key]['ElasticNet'][tmpKey].update(tmpDict[tmpKey])
mp.plotLinearModelResults(dataSets,
resultDict,