diff options
| -rw-r--r-- | simulation/plot_utils.py | 26 | ||||
| -rw-r--r-- | simulation/utils_blocks.py | 35 | ||||
| -rw-r--r-- | simulation/vi_blocks.py | 4 |
3 files changed, 62 insertions, 3 deletions
diff --git a/simulation/plot_utils.py b/simulation/plot_utils.py new file mode 100644 index 0000000..af5269c --- /dev/null +++ b/simulation/plot_utils.py @@ -0,0 +1,26 @@ +import matplotlib.pyplot as plt +import argparse +import json +import seaborn +seaborn.set_style('whitegrid') + +parser = argparse.ArgumentParser(description='Process logs') +parser.add_argument('-x', help='name of parameters on x axis', default='time') +parser.add_argument('-y', help='name of parameters on y axis', default='rmse') +parser.add_argument('f', help='list of logs to parse', nargs='+') +parser.add_argument('-dest', help='name of figure to save', default='fig.png') +args = parser.parse_args() + +for file_name in args.f: + x, y = [], [] + with open(file_name) as f: + for line in f: + jason = json.loads(line) + x.append(jason[args.x]) + y.append(jason[args.y]) + plt.plot(x, y, label=file_name) + +plt.legend() +plt.xlabel(args.x) +plt.ylabel(args.y) +plt.savefig(args.dest) diff --git a/simulation/utils_blocks.py b/simulation/utils_blocks.py index 5e91658..0d30786 100644 --- a/simulation/utils_blocks.py +++ b/simulation/utils_blocks.py @@ -38,7 +38,9 @@ class ActiveLearning(be.SimpleExtension): def do(self, which_callback, *args): out_degree = np.sum(self.dataset.graph, axis=1) self.dataset.node_p = out_degree / np.sum(out_degree) - print(self.dataset.node_p) + +# def do(self, which_callback, *args): + class JSONDump(be.SimpleExtension): @@ -119,3 +121,34 @@ def dynamic_data_stream(graph, batch_size): data_set = LearnedDataset(node_p, graph) scheme = fuel.schemes.ConstantScheme(batch_size) return fuel.streams.DataStream(dataset=data_set, iteration_scheme=scheme) + + +if __name__ == "__main__": + batch_size = 100 + n_obs = 1000 + frequency = 1 + graph = utils.create_wheel(1000) + print('GRAPH:\n', graph, '\n-------------\n') + + g_shared = theano.shared(value=graph, name='graph') + x, s, params, cost = create_mle_model(graph) + rmse = rmse_error(g_shared, params) + error = relative_error(g_shared, params) + + alg = algorithms.GradientDescent( + cost=-cost, parameters=[params], step_rule=blocks.algorithms.AdaDelta() + ) + data_stream = create_learned_data_stream(graph, batch_size) + #data_stream = create_fixed_data_stream(n_obs, graph, batch_size) + loop = main_loop.MainLoop( + alg, data_stream, + extensions=[ + be.FinishAfter(after_n_batches=10**4), + bm.TrainingDataMonitoring([cost, rmse, error], + every_n_batches=frequency), + be.Printing(every_n_batches=frequency), + JSONDump("tmpactive_log.json", every_n_batches=frequency), + ActiveLearning(data_stream.dataset, every_n_batches=frequency) + ], + ) + loop.run() diff --git a/simulation/vi_blocks.py b/simulation/vi_blocks.py index 94038a5..1177979 100644 --- a/simulation/vi_blocks.py +++ b/simulation/vi_blocks.py @@ -51,8 +51,8 @@ def create_vi_model(n_nodes, n_samp=100): if __name__ == "__main__": - n_cascades = 10000 - batch_size = 1000 + #n_cascades = 10000 + batch_size = 10 n_samples = 50 graph = utils.create_random_graph(n_nodes=4) print('GRAPH:\n', graph, '\n-------------\n') |