diff options
Diffstat (limited to 'simulation')
| -rw-r--r-- | simulation/mle_blocks.py | 4 | ||||
| -rw-r--r-- | simulation/utils_blocks.py | 8 | ||||
| -rw-r--r-- | simulation/vi_blocks.py | 15 |
3 files changed, 15 insertions, 12 deletions
diff --git a/simulation/mle_blocks.py b/simulation/mle_blocks.py index 98bc257..1b1cf4d 100644 --- a/simulation/mle_blocks.py +++ b/simulation/mle_blocks.py @@ -32,14 +32,14 @@ if __name__ == "__main__": batch_size = 100 #n_obs = 100000 freq = 10 - graph = utils.create_wheel(1000) + graph = utils.create_wheel(100) print('GRAPH:\n', graph, '\n-------------\n') g_shared = theano.shared(value=graph, name='graph') x, s, params, cost = create_mle_model(graph) rmse = ub.rmse_error(g_shared, params) - error = ub.relative_error(g_shared, params) + error = ub.absolute_error(g_shared, params) alg = algorithms.GradientDescent( cost=-cost, parameters=[params], step_rule=algorithms.AdaDelta() diff --git a/simulation/utils_blocks.py b/simulation/utils_blocks.py index 00b429e..2dc9f85 100644 --- a/simulation/utils_blocks.py +++ b/simulation/utils_blocks.py @@ -85,18 +85,18 @@ def rmse_error(graph, params): diff = (graph - params) ** 2 subarray = tsr.arange(n_nodes) tsr.set_subtensor(diff[subarray, subarray], 0) - rmse = tsr.sum(diff) / (n_nodes ** 2) + rmse = tsr.sqrt(tsr.sum(diff) / (n_nodes ** 2)) rmse.name = 'rmse' return rmse -def relative_error(graph, params): +def absolute_error(graph, params): n_nodes = graph.shape[0] diff = abs(graph - params) subarray = tsr.arange(n_nodes) tsr.set_subtensor(diff[subarray, subarray], 0) - error = tsr.sum(tsr.switch(tsr.eq(graph, 0.), 0., diff / graph)) / n_nodes - error.name = 'rel_error' + error = tsr.sum(diff) / (n_nodes ** 2) + error.name = 'abs_error' return error diff --git a/simulation/vi_blocks.py b/simulation/vi_blocks.py index 50c7fb1..e2e3bd9 100644 --- a/simulation/vi_blocks.py +++ b/simulation/vi_blocks.py @@ -26,7 +26,7 @@ def create_vi_model(n_nodes, n_samp=100): """return variational inference theano computation graph""" def aux(a, b): rand = a + b * np.random.normal(size=(n_nodes, n_nodes)) - return np.clip(rand, 1e-3, 1 - 1e-3).astype(theano.config.floatX) + return np.clip(rand, 1e-10, 1000).astype(theano.config.floatX) x = tsr.matrix(name='x', dtype='int8') s = tsr.matrix(name='s', dtype='int8') @@ -54,12 +54,14 @@ if __name__ == "__main__": batch_size = 100 freq = 10 graph = utils.create_wheel(1000) + g_shared = theano.shared(value=graph, name='graph') n_samples = 50 - #graph = utils.create_random_graph(n_nodes=10) + # graph = utils.create_random_graph(n_nodes=10) print('GRAPH:\n', graph, '\n-------------\n') x, s, mu, sig, cost = create_vi_model(len(graph), n_samples) - rmse = ub.rmse_error(graph, mu) + rmse = ub.rmse_error(g_shared, mu) + error = ub.absolute_error(g_shared, mu) step_rules = algorithms.CompositeRule([algorithms.AdaDelta(), ClippedParams(1e-3, 1000)]) @@ -72,11 +74,12 @@ if __name__ == "__main__": log_backend="sqlite", extensions=[ be.FinishAfter(after_n_batches=10**3), - bm.TrainingDataMonitoring([cost, rmse, mu], every_n_batches=freq), + bm.TrainingDataMonitoring([cost, rmse, mu, error], + every_n_batches=freq), be.Printing(every_n_batches=freq, after_epoch=False), ub.JSONDump("logs/nonactive_vi.json", every_n_batches=freq), - #ub.ActiveLearning(dataset=data_stream.dataset, params=graph, - #every_n_batches=freq) + # ub.ActiveLearning(dataset=data_stream.dataset, params=graph, + # every_n_batches=freq) ] ) loop.run() |