# -*- coding: UTF-8 -*-
# File: symbolic_functions.py
# Author: Yuxin Wu <ppwwyyxx@gmail.com>
import tensorflow as tf
import numpy as np
from ..utils import logger
def prediction_incorrect(logits, label, topk=1):
"""
:param logits: NxC
:param label: N
:returns: a float32 vector of length N with 0/1 values, 1 meaning incorrect prediction
"""
return tf.cast(tf.logical_not(tf.nn.in_top_k(logits, label, topk)), tf.float32)
def flatten(x):
"""
Flatten the tensor.
"""
return tf.reshape(x, [-1])
def batch_flatten(x):
"""
Flatten the tensor except the first dimension.
"""
shape = x.get_shape().as_list()[1:]
if None not in shape:
return tf.reshape(x, [-1, np.prod(shape)])
return tf.reshape(x, tf.pack([tf.shape(x)[0], -1]))
def logSoftmax(x):
"""
Batch log softmax.
:param x: NxC tensor.
:returns: NxC tensor.
"""
logger.warn("symbf.logSoftmax is deprecated in favor of tf.nn.log_softmax")
return tf.nn.log_softmax(x)
def class_balanced_binary_class_cross_entropy(pred, label, name='cross_entropy_loss'):
"""
The class-balanced cross entropy loss for binary classification,
as in `Holistically-Nested Edge Detection
<http://arxiv.org/abs/1504.06375>`_.
:param pred: size: b x ANYTHING. the predictions in [0,1].
:param label: size: b x ANYTHING. the ground truth in {0,1}.
:returns: class-balanced binary classification cross entropy loss
"""
z = batch_flatten(pred)
y = tf.cast(batch_flatten(label), tf.float32)
count_neg = tf.reduce_sum(1. - y)
count_pos = tf.reduce_sum(y)
beta = count_neg / (count_neg + count_pos)
eps = 1e-8
loss_pos = -beta * tf.reduce_mean(y * tf.log(tf.abs(z) + eps), 1)
loss_neg = (1. - beta) * tf.reduce_mean((1. - y) * tf.log(tf.abs(1. - z) + eps), 1)
cost = tf.sub(loss_pos, loss_neg)
cost = tf.reduce_mean(cost, name=name)
return cost
def print_stat(x, message=None):
""" a simple print op.
Use it like: x = print_stat(x)
"""
if message is None:
message = x.op.name
return tf.Print(x, [tf.reduce_mean(x), x], summarize=20,
message=message, name='print_' + x.op.name)
def rms(x, name=None):
if name is None:
name = x.op.name + '/rms'
with tf.name_scope(None): # name already contains the scope
return tf.sqrt(tf.reduce_mean(tf.square(x)), name=name)
return tf.sqrt(tf.reduce_mean(tf.square(x)), name=name)
def huber_loss(x, delta=1, name=None):
if name is None:
name = 'huber_loss'
sqrcost = tf.square(x)
abscost = tf.abs(x)
return tf.reduce_sum(
tf.select(abscost < delta,
sqrcost * 0.5,
abscost * delta - 0.5 * delta ** 2),
name=name)
def get_scalar_var(name, init_value):
return tf.get_variable(name, shape=[],
initializer=tf.constant_initializer(init_value),
trainable=False)
