taxi

Winning entry to the Kaggle taxi competition
git clone https://esimon.eu/repos/taxi.git
Log | Files | Refs | README

rnn_tgtcls_1.py (954B)


      1 import os
      2 import cPickle
      3 
      4 from blocks.initialization import IsotropicGaussian, Constant
      5 
      6 import data
      7 from model.rnn_tgtcls import Model, Stream
      8 
      9 class EmbedderConfig(object):
     10     __slots__ = ('dim_embeddings', 'embed_weights_init')
     11 
     12 pre_embedder = EmbedderConfig()
     13 pre_embedder.embed_weights_init = IsotropicGaussian(0.001)
     14 pre_embedder.dim_embeddings = [ 
     15     ('week_of_year', 52, 10),
     16     ('day_of_week', 7, 10),
     17     ('qhour_of_day', 24 * 4, 10),
     18     ('day_type', 3, 10),
     19     ('taxi_id', 448, 10),
     20 ]
     21 
     22 post_embedder = EmbedderConfig()
     23 post_embedder.embed_weights_init = IsotropicGaussian(0.001)
     24 post_embedder.dim_embeddings = [ 
     25     ('origin_call', data.origin_call_train_size, 10), 
     26     ('origin_stand', data.stands_size, 10),
     27 ]
     28 
     29 with open(os.path.join(data.path, 'arrival-clusters.pkl')) as f: tgtcls = cPickle.load(f)
     30 
     31 hidden_state_dim = 100 
     32 weights_init = IsotropicGaussian(0.01)
     33 biases_init = Constant(0.001)
     34 
     35 batch_size = 10
     36 batch_sort_size = 10