/
tilt-gpfa.py
204 lines (187 loc) · 7.35 KB
/
tilt-gpfa.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
# tilt-gpfa.py
# perform gpfa on tilt dataset
pgpfa = __import__("poisson-gpfa")
util = pgpfa.funs.util
dm = pgpfa.funs.datamanager
engine = pgpfa.funs.engine
import matplotlib.pyplot as plt
import numpy as np
import json
import dill
import sys
def load_data(binSize,binFilename,classifierFilename):
# load binned resp data
with open(binFilename) as json_file:
data=json.load(json_file)
# load classified trial indices
with open(classifierFilename) as json_file:
data_classifier=json.load(json_file)
# get correct trials for given binSize
binStr='bin_{}ms'.format(binSize)
correct_trials = data_classifier[binStr]['correct_trials']
return [correct_trials, data]
def iter_gpfa(data_in,latent_dims,params,fig_params):
# -> iterate through different dimensions for latent space
# unpack params
ydim = params['ydim']
trialDur = params['trialDur']
binSize = params['binSize']
maxEMiter = params['maxEMiter']
dOffset = params['dOffset']
infType = params['infType']
numTrials = params['numTrials']
# unpack fig_params
preamble = fig_params['preamble']
# generate tilt dataset obj
tilt_set = dm.TiltDataset(
data = data_in,
ydim = ydim,
trialDur = trialDur,
binSize = binSize,
numTrials = numTrials,
numTrData = True
)
# iterate through # of latent dimensions
out = dict()
for xdim in latent_dims:
print('-> Running gpfa for xdim={}'.format(xdim))
initParams = util.initializeParams(xdim, ydim, tilt_set)
if infType=='batch':
# Fit using vanilla (batch) EM; longer runtime
fitBatch = engine.PPGPFAfit(
experiment = tilt_set,
initParams = initParams,
inferenceMethod = 'laplace',
EMmode = 'Batch',
maxEMiter = maxEMiter)
out[xdim]=fitBatch
elif infType=='online':
# Fit using online EM; shorter runtime
fitOnline = engine.PPGPFAfit(
experiment = tilt_set,
initParams = initParams,
EMmode = 'Online',
maxEMiter = maxEMiter,
inferenceMethod = 'laplace',
batchSize = 5)
# plot, save figs
fitOnline.plotParamSeq();
plt.savefig('{}-paramSeq-xdim{}.png'.format(preamble,xdim))
plt.show()
fitOnline.plotTrajectory(1);
plt.savefig('{}-latentTraj-xdim{}.png'.format(preamble,xdim))
plt.show()
out[xdim]=fitOnline
return out
def process_tilt_data(binSize,correct_trials,data,latent_dims):
# -> data = binned json data to be formatted for gpfa
# -> latent_dims = array of ints for different latent trajectory sizes
fitObj = dict()
i_trial = 0
for event_name, event_data in data[str(binSize)]['data'].items():
print('event: {}'.format(event_name))
data_in=[]
for trial_name, trial_data in event_data.items():
if i_trial in correct_trials:
data_in.append({'Y':np.asarray(trial_data['Y'])})
i_trial = i_trial + 1
numTrials = len(data_in)
data_sub=data['5']
# Specify dataset & fitting parameters
ydim = data_in[0]['Y'][:,1].size # neurons
trialDur = data_sub['trialDur'] # in ms
binSize = data_sub['binSize'] # in ms
maxEMiter = 100 # expectation maximization iterations
dOffset = 1 # controls firing rate
params={
'ydim':ydim,
'trialDur':trialDur,
'binSize':binSize,
'maxEMiter':maxEMiter,
'dOffset':dOffset,
'infType':'batch',
'numTrials': numTrials
}
filepath = "images/"
trial_type = "allTrials"
preamble = "{}{}-{}".format(filepath,event_name,trial_type)
fig_params={
'preamble': preamble
}
fitObj[event_name]=iter_gpfa(data_in,latent_dims,params,fig_params)
return fitObj
def plot_3d_traj(filepath,trial_type,preamble):
# plot multiple latent trajectories
filepath = "images/"
trial_type = "fitBatch5_correctTrialsAndEst"
preamble = "{}{}".format(filepath,trial_type)
# trial_i = 0
for event_name, event_data in fitObj.items():
fig = plt.figure(figsize=(5,5))
ax = fig.gca(projection='3d')
fitObjTemp=event_data[3]
print('event: {}'.format(event_name))
# print(len(fitObjTemp.infRes['post_mean']))
# print(fitObjTemp.infRes['post_mean'])
for traj in fitObjTemp.infRes['post_mean']:
# traj = fitObj.infRes['post_mean']
# if trial_i in correct_trials:
ax.plot3D(traj[0],traj[1],traj[2])
# trial_i = trial_i + 1
plt.title('{} Latent Trajectory'.format(event_name));
ax.set_xlabel('xdim1')
ax.set_ylabel('xdim2')
ax.set_zlabel('xdim3')
# plt.show();
plt.savefig('{}-{}-latentTraj-3D.png'.format(preamble,event_name))
def plot_trial_traj(filepath,correct_trials,dim,fitObj):
# filepath = 'C:/Users/Mason/Box Sync/UC Davis/00 Quarters/Spring 2019/MAE298/03 Project/poisson-gpfa/images/classifiedTrialsAndEst/latentTraj'
incr=1
# dim=3
last_i = 0
for event_name, event_data in fitObj.items():
num_trials = len(event_data[dim].infRes['post_mean'])
print('num_trials: {}'.format(num_trials))
for i in range(0,numTrials+1):
trial_num = correct_trials[last_i+i]
# print("plotting trial #{:03d}".format(trial_num))
title_str = 'Trial #{:03d}'.format(trial_num)
save_str = filepath+'\{}-trial{:03d}-{}traj'.format(event_name,trial_num,dim)
fitObj[event_name][dim].plotTrajectory(i);plt.title(title_str);plt.savefig(save_str);
# plt.show();
plt.close();
last_i = last_i + num_trials
print('last_i: {}'.format(last_i))
if __name__ == '__main__':
# sys args
n_args = len(sys.argv)
if n_args == 2:
min_dim = 3
max_dim = int(sys.argv[1])
incr = 2
elif n_args == 3:
min_dim = int(sys.argv[1])
max_dim = int(sys.argv[2])
incr = 2
elif n_args == 4:
min_dim = int(sys.argv[1])
max_dim = int(sys.argv[2])
incr = int(sys.argv[3])
else:
# wrong number of sys args; use predefined range
min_dim = 3
max_dim = 11
incr = 2
latent_dims = list(range(min_dim,max_dim,incr))
print("Latent Dimensions to infer: {}".format(latent_dims))
# load data
print("Loading data...")
binSize=5
binFilename='data/gpfa_inputs.json'
classifierFilename='data/tilt_psthclassifier.json'
[correct_trials, data]=load_data(binSize,binFilename,classifierFilename)
# perform gpfa for different latent trajectory dimensionalities
print("Begin gpfa iteration...")
fitObj=process_tilt_data(binSize,correct_trials,data,latent_dims)
# dump data for local post-process
dill.dump_session('fitBatch_{}to{}by{}.db'.format(min_dim,max_dim,incr))