TF-Ranking - Portfolio

tf_portfolio_ranking.py

import tensorflow as tf
import tensorflow_ranking as tfr

_TRAIN_DATA_PATH="/data/train.txt"
_TEST_DATA_PATH="/data/test.txt"
_LOSS="approx_ndcg_loss"
_N_ASSETS=100
_N_FEATURES=16
_BATCH_SIZE=32
_HIDDEN_LAYER_DIMS=["20", "10"]


# Input Pipeline
def input_fn(path):
    data = tf.data.Dataset.from_generator(
        tfr.data.libsvm_generator(path, _N_FEATURES, _N_ASSETS),
        output_types=({str(k): tf.float32 for k in range(1,_N_FEATURES+1)}, tf.float32),
        output_shapes=(
            {str(k): tf.TensorShape([_N_ASSETS, 1]) for k in range(1,_N_FEATURES+1)},
            tf.TensorShape([_N_ASSETS])
        )
    )

    data = data.shuffle(1000).repeat().batch(_BATCH_SIZE)
  
    return data.make_one_shot_iterator().get_next()


def example_feature_columns():
    """Returns the example feature columns."""
    
    feature_names = ["%d" % (i + 1) for i in range(0, _N_FEATURES)]
    
    return {name: tf.feature_column.numeric_column(name, shape=(1,), default_value=0.0)
            for name in feature_names}


# Scoring Function
def make_score_fn():
    """Returns a scoring function to build `EstimatorSpec`."""

    def _score_fn(context_features, group_features, mode, params, config):
        """Defines the network to score assets."""
        
        del params
        del config
        # Define input layer.
        example_input = [
            tf.layers.flatten(group_features[name])
            for name in sorted(example_feature_columns())
        ]
        input_layer = tf.concat(example_input, 1)

        cur_layer = input_layer
        for i, layer_width in enumerate(int(d) for d in _HIDDEN_LAYER_DIMS):
            cur_layer = tf.layers.dense(
                cur_layer,
                units=layer_width,
                activation="tanh")

        logits = tf.layers.dense(cur_layer, units=1)
        
        return logits

    return _score_fn


# Evaluation Metric
def eval_metric_fns():
      """Returns a dict from name to metric functions.
      Returns:
        A dict mapping from metric name to a metric function with above signature.
      """
      
    metric_fns = {}
    metric_fns.update({
        "metric/ndcg@%d" % topn: tfr.metrics.make_ranking_metric_fn(
            tfr.metrics.RankingMetricKey.NDCG, topn=topn)
        for topn in [1, 3, 5, 10]
    })
    
    return metric_fns


# Estimator
def get_estimator(hparams):
    """Create a ranking estimator.
    Args:
    hparams: (tf.contrib.training.HParams) a hyperparameters object.
    Returns:
    tf.learn `Estimator`.
    """
    
    def _train_op_fn(loss):
    """Defines train op used in ranking head."""
    return tf.contrib.layers.optimize_loss(
        loss=loss,
        global_step=tf.train.get_global_step(),
        learning_rate=hparams.learning_rate,
        optimizer="Adagrad")

    ranking_head = tfr.head.create_ranking_head(
        loss_fn=tfr.losses.make_loss_fn(_LOSS),
        eval_metric_fns=eval_metric_fns(),
        train_op_fn=_train_op_fn)

    return tf.estimator.Estimator(
        model_fn=tfr.model.make_groupwise_ranking_fn(
          group_score_fn=make_score_fn(),
          group_size=1,
          transform_fn=None,
          ranking_head=ranking_head),
        params=hparams)


# Initialize estimator
hparams = tf.contrib.training.HParams(learning_rate=0.05)
ranker = get_estimator(hparams)

# Train model
ranker.train(input_fn=lambda: input_fn(_TRAIN_DATA_PATH), steps=100)

# Evaluate model
ranker.evaluate(input_fn=lambda: input_fn(_TEST_DATA_PATH), steps=100)

# Visualize
ranker.model_dir

Hi Luis: Can you share the txt files that you are training the model with? I'd like to reproduce the results. Thanks Ram

Hola Luis, podrías compartir los archivos de entrenamiento y testing.

De antemano muchas gracias!

Hello Luis, can you share a sample of your training file or format? Thank you.

Hi, could you share your data with us? Thanks

Hi, could you share just some snippets of the data (can be anonymized) just to be able to run your code on a working example? Thanks

+1 for sharing the data or at least its format