active learning for mle + variational inf. (not bug-free)

1 files changed, 0 insertions, 110 deletions

diff --git a/simulation/vi_theano.py b/simulation/vi_theano.py
deleted file mode 100644
index cfd1dcf..0000000
--- a/simulation/vi_theano.py
+++ /dev/null
@@ -1,110 +0,0 @@
-import main as mn
-import theano
-from theano import tensor as tsr
-import theano.tensor.shared_randomstreams
-import numpy as np
-
-n_cascades = 1000
-n_nodes = 3
-n_samples = 100
-srng = tsr.shared_randomstreams.RandomStreams(seed=123)
-lr = 1e-1
-n_epochs = 1
-
-###############Variational Inference####################
-
-# Declare Theano variables
-mu = theano.shared(.5 + .2 * np.random.normal(size=(1, n_nodes, n_nodes)),
-                    name="mu", broadcastable=(True, False, False))
-sig = theano.shared(.1 + .04 * np.random.normal(size=(1, n_nodes, n_nodes)),
-                    name="sig", broadcastable=(True, False, False))
-mu0 = theano.shared(.5 + .2 * np.random.normal(size=(1, n_nodes, n_nodes)),
-                    name="mu", broadcastable=(True, False, False))
-sig0 = theano.shared(.1 + .04 * np.random.normal(size=(1, n_nodes, n_nodes)),
-                    name="sig", broadcastable=(True, False, False))
-x = tsr.matrix(name='x', dtype='int8')
-s = tsr.matrix(name='s', dtype='int8')
-
-# Construct Theano graph
-theta = srng.normal((n_samples, n_nodes, n_nodes)) * sig + mu
-y = tsr.maximum(tsr.dot(x, theta), 1e-3)
-infect = tsr.log(1. - tsr.exp(-y[0:-1])).dimshuffle(1, 0, 2)
-lkl_pos = tsr.sum(infect * (x[1:] & s[1:])) / n_samples
-lkl_neg = tsr.sum(-y[0:-1].dimshuffle(1, 0, 2) * (~x[1:] & s[1:])) / n_samples
-lkl = lkl_pos + lkl_neg
-kl = tsr.sum(tsr.log(sig / sig0) + (sig0**2 + (mu0 - mu)**2)/(2*sig)**2)
-res = lkl + kl
-
-gmu, gsig = theano.gradient.grad(lkl, [mu, sig])
-gmukl, gsigkl = theano.grad(kl, [mu, sig])
-
-# Compile into functions
-loglkl_full = theano.function([x, s], lkl)
-train = theano.function(inputs=[x, s], outputs=res,
-                        updates=((mu, tsr.clip(mu + lr * gmu, 0, 1)),
-                                 (sig, tsr.clip(sig + lr * gsig, 1e-3, 1))))
-train_kl = theano.function(inputs=[], outputs=[],
-                           updates=((mu, tsr.clip(mu + lr * gmukl, 0, 1)),
-                                   (sig, tsr.clip(sig + lr * gsigkl, 1e-3, 1))))
-
-
-###############Maximum Likelihood#####################
-
-x = tsr.matrix(name='x', dtype='int8')
-s = tsr.matrix(name='s', dtype='int8')
-params = theano.shared(.5 + .01*np.random.normal(size=(n_nodes, n_nodes)),
-                        name='params')
-y = tsr.maximum(tsr.dot(x, params), 1e-5)
-infect = tsr.log(1. - tsr.exp(-y[0:-1]))
-lkl_pos = tsr.sum(infect * (x[1:] & s[1:]))
-lkl_neg = tsr.sum(-y[0:-1] * (~x[1:] & s[1:]))
-lkl_mle = lkl_pos + lkl_neg
-gparams = theano.gradient.grad(lkl_mle, params)
-train_mle = theano.function(inputs=[x, s], outputs=lkl_mle, updates=[(params,
-                                        tsr.clip(params + lr * gparams, 0, 1))])
-
-
-if __name__ == "__main__":
-    graph = .5 * np.random.binomial(2, p=.5, size=(n_nodes, n_nodes))
-    for k in range(len(graph)):
-        graph[k, k] = 0
-    p = 0.5
-    graph = np.log(1. / (1 - p * graph))
-
-    graph = np.array([[0, 0, 1], [0, 0, 0.5], [0, 0, 0]])
-    p = 0.5
-    graph = np.log(1. / (1 - p * graph))
-
-    cascades = mn.build_cascade_list(mn.simulate_cascades(n_cascades, graph),
-                                     collapse=True)
-    x_obs, s_obs = cascades[0], cascades[1]
-
-    #mle
-    lkl_plot = []
-    if 1:
-        for i in range(n_epochs):
-            for k in xrange(len(x_obs)/100):
-                xt = x_obs[k*100:(k+1)*100]
-                st = s_obs[k*100:(k+1)*100]
-                lkl = train_mle(xt, st)
-                lkl_plot.append(lkl)
-                print(params.get_value())
-        print(graph)
-        w = params.get_value()
-        for k in range(len(w)):
-            w[k, k] = 0
-        print(w)
-        #import matplotlib.pyplot as plt
-        #plt.plot(lkl_plot)
-        #plt.show()
-
-    #variational inference
-    if 0:
-        for i in range(n_epochs):
-            train_kl()
-            for k in xrange(len(x_obs)/100):
-                cost = train(x_obs[k*100:(k+1)*100], s_obs[k*100:(k+1)*100])
-            print(cost)
-        print(graph)
-        print(mu.get_value())
-        print(sig.get_value())