taxi

commit 66d2717188e189fde5422576740903ca8e488f63
parent 6cf80aff3e4dc57d13a33c2946bc0ae57cfae6b8
Author: Alex Auvolat <alex.auvolat@ens.fr>
Date:   Thu,  2 Jul 2015 15:00:51 -0400

Add small models

Diffstat:
M
config/dest_mlp_tgtcls_1_cswdtx_alexandre.py
 | 
5
+++--
A
config/dest_mlp_tgtcls_1_cswdtx_small.py
 | 
38
++++++++++++++++++++++++++++++++++++++
A
config/dest_mlp_tgtcls_2_cswdtx_small.py
 | 
38
++++++++++++++++++++++++++++++++++++++

3 files changed, 79 insertions(+), 2 deletions(-)
diff --git a/config/dest_mlp_tgtcls_1_cswdtx_alexandre.py b/config/dest_mlp_tgtcls_1_cswdtx_alexandre.py
@@ -2,6 +2,7 @@ import os
 import cPickle
 
 from blocks.initialization import IsotropicGaussian, Constant
+from blocks.algorithms import Momentum
 
 import data
 from model.dest_mlp_tgtcls import Model, Stream
@@ -29,8 +30,8 @@ embed_weights_init = IsotropicGaussian(0.01)
 mlp_weights_init = IsotropicGaussian(0.1) 
 mlp_biases_init = Constant(0.01)
 
-learning_rate = 0.01
-momentum = 0.9
+step_rule = Momentum(learning_rate=0.01, momentum=0.9)
+
 batch_size = 200
 
 valid_set = 'cuts/test_times_0'
diff --git a/config/dest_mlp_tgtcls_1_cswdtx_small.py b/config/dest_mlp_tgtcls_1_cswdtx_small.py
@@ -0,0 +1,38 @@
+import os
+import cPickle
+
+from blocks.initialization import IsotropicGaussian, Constant
+from blocks.algorithms import Momentum
+
+import data
+from model.dest_mlp_tgtcls import Model, Stream
+
+
+n_begin_end_pts = 5     # how many points we consider at the beginning and end of the known trajectory
+
+with open(os.path.join(data.path, 'arrival-clusters.pkl')) as f: tgtcls = cPickle.load(f)
+
+dim_embeddings = [
+    ('origin_call', data.origin_call_train_size, 10),
+    ('origin_stand', data.stands_size, 10),
+    ('week_of_year', 52, 10),
+    ('day_of_week', 7, 10),
+    ('qhour_of_day', 24 * 4, 10),
+    ('day_type', 3, 10),
+    ('taxi_id', 448, 10),
+]
+
+dim_input = n_begin_end_pts * 2 * 2 + sum(x for (_, _, x) in dim_embeddings)
+dim_hidden = [100]
+dim_output = tgtcls.shape[0]
+
+embed_weights_init = IsotropicGaussian(0.01)
+mlp_weights_init = IsotropicGaussian(0.1) 
+mlp_biases_init = Constant(0.01)
+
+step_rule = Momentum(learning_rate=0.01, momentum=0.9)
+
+batch_size = 200
+
+valid_set = 'cuts/test_times_0'
+max_splits = 100
diff --git a/config/dest_mlp_tgtcls_2_cswdtx_small.py b/config/dest_mlp_tgtcls_2_cswdtx_small.py
@@ -0,0 +1,38 @@
+import os
+import cPickle
+
+from blocks.initialization import IsotropicGaussian, Constant
+from blocks.algorithms import Momentum
+
+import data
+from model.dest_mlp_tgtcls import Model, Stream
+
+
+n_begin_end_pts = 5     # how many points we consider at the beginning and end of the known trajectory
+
+with open(os.path.join(data.path, 'arrival-clusters.pkl')) as f: tgtcls = cPickle.load(f)
+
+dim_embeddings = [
+    ('origin_call', data.origin_call_train_size, 10),
+    ('origin_stand', data.stands_size, 10),
+    ('week_of_year', 52, 10),
+    ('day_of_week', 7, 10),
+    ('qhour_of_day', 24 * 4, 10),
+    ('day_type', 3, 10),
+    ('taxi_id', 448, 10),
+]
+
+dim_input = n_begin_end_pts * 2 * 2 + sum(x for (_, _, x) in dim_embeddings)
+dim_hidden = [100, 100]
+dim_output = tgtcls.shape[0]
+
+embed_weights_init = IsotropicGaussian(0.01)
+mlp_weights_init = IsotropicGaussian(0.1) 
+mlp_biases_init = Constant(0.01)
+
+step_rule = Momentum(learning_rate=0.01, momentum=0.9)
+
+batch_size = 200
+
+valid_set = 'cuts/test_times_0'
+max_splits = 100