ó ¾÷Xc@@s|ddlmZddlmZddlmZddlZddl m Z defd„ƒYZ d efd „ƒYZ dS( i(tabsolute_importi(tLayer(tbackendN(t interfacest GaussianNoisecB@s5eZdZejd„ƒZdd„Zd„ZRS(s“Apply additive zero-centered Gaussian noise. This is useful to mitigate overfitting (you could see it as a form of random data augmentation). Gaussian Noise (GS) is a natural choice as corruption process for real valued inputs. As it is a regularization layer, it is only active at training time. # Arguments stddev: float, standard deviation of the noise distribution. # Input shape Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model. # Output shape Same shape as input. cK@s,tt|ƒj|t|_||_dS(N(tsuperRt__init__tTruetsupports_maskingtstddev(tselfR tkwargs((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyR s c@s(‡‡fd†}tj|ˆd|ƒS(Nc@s,ˆtjdtjˆƒdddˆjƒS(NtshapetmeangR (tKt random_normalR R ((tinputsR (s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pytnoised'sttraining(Rtin_train_phase(R RRR((RR s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pytcall&scC@sKi|jd6}tt|ƒjƒ}tt|jƒƒt|jƒƒƒS(NR (R RRt get_configtdicttlisttitems(R tconfigt base_config((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyR-sN( t__name__t __module__t__doc__Rtlegacy_gaussiannoise_supportRtNoneRR(((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyR s tGaussianDropoutcB@s5eZdZejd„ƒZdd„Zd„ZRS(sëApply multiplicative 1-centered Gaussian noise. As it is a regularization layer, it is only active at training time. # Arguments rate: float, drop probability (as with `Dropout`). The multiplicative noise will have standard deviation `sqrt(rate / (1 - rate))`. # Input shape Arbitrary. Use the keyword argument `input_shape` (tuple of integers, does not include the samples axis) when using this layer as the first layer in a model. # Output shape Same shape as input. # References - [Dropout: A Simple Way to Prevent Neural Networks from Overfitting Srivastava, Hinton, et al. 2014](http://www.cs.toronto.edu/~rsalakhu/papers/srivastava14a.pdf) cK@s,tt|ƒj|t|_||_dS(N(RR RRRtrate(R R!R ((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyRIs c@sKdˆjkodknrG‡‡fd†}tj|ˆd|ƒSˆS(Niic@sFtjˆjdˆjƒ}ˆtjdtjˆƒddd|ƒS(Ngð?R R R (tnptsqrtR!RRR (R (RR (s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyRQsR(R!RR(R RRR((RR s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyROscC@sKi|jd6}tt|ƒjƒ}tt|jƒƒt|jƒƒƒS(NR!(R!RR RRRR(R RR((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyRYsN( RRRRtlegacy_gaussiandropout_supportRRRR(((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyR 3s ( t __future__RtengineRtRRtnumpyR"tlegacyRRR (((s1/tmp/pip-build-isqEY4/keras/keras/layers/noise.pyts  )