σ ΎχXc@@sddlmZddlZddlmZddlmZddlmZde fd„ƒYZ d e fd „ƒYZ d e fd „ƒYZ d e fd„ƒYZ de fd„ƒYZe Ze Ze ZeZeZeZeZd„Zdd„Zd„ZdS(i(tabsolute_importNi(tbackend(tserialize_keras_object(tdeserialize_keras_objectt ConstraintcB@seZd„Zd„ZRS(cC@s|S(N((tselftw((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyt__call__ scC@siS(N((R((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyt get_config s(t__name__t __module__RR(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRs tMaxNormcB@s/eZdZddd„Zd„Zd„ZRS(sκMaxNorm weight constraint. Constrains the weights incident to each hidden unit to have a norm less than or equal to a desired value. # Arguments m: the maximum norm for the incoming weights. axis: integer, axis along which to calculate weight norms. For instance, in a `Dense` layer the weight matrix has shape `(input_dim, output_dim)`, set `axis` to `0` to constrain each weight vector of length `(input_dim,)`. In a `Convolution2D` layer with `data_format="channels_last"`, the weight tensor has shape `(rows, cols, input_depth, output_depth)`, set `axis` to `[0, 1, 2]` to constrain the weights of each filter tensor of size `(rows, cols, input_depth)`. # 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) iicC@s||_||_dS(N(t max_valuetaxis(RR R ((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyt__init__)s cC@sdtjtjtj|ƒd|jdtƒƒ}tj|d|jƒ}||tjƒ|9}|S(NR tkeepdimsi( tKtsqrttsumtsquareR tTruetclipR tepsilon(RRtnormstdesired((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR-s0cC@si|jd6|jd6S(NR R (R R (R((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR3s (R R t__doc__RRR(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR s tNonNegcB@seZdZd„ZRS(s/Constrains the weights to be non-negative. cC@s&|tj|dktjƒƒ9}|S(Ng(Rtcasttfloatx(RR((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR<s"(R R RR(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR8stUnitNormcB@s,eZdZdd„Zd„Zd„ZRS(s―Constrains the weights incident to each hidden unit to have unit norm. # Arguments axis: integer, axis along which to calculate weight norms. For instance, in a `Dense` layer the weight matrix has shape `(input_dim, output_dim)`, set `axis` to `0` to constrain each weight vector of length `(input_dim,)`. In a `Convolution2D` layer with `data_format="channels_last"`, the weight tensor has shape `(rows, cols, input_depth, output_depth)`, set `axis` to `[0, 1, 2]` to constrain the weights of each filter tensor of size `(rows, cols, input_depth)`. icC@s ||_dS(N(R (RR ((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRRsc C@s<|tjƒtjtjtj|ƒd|jdtƒƒS(NR R(RRRRRR R(RR((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRUs' cC@si|jd6S(NR (R (R((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRZs(R R RRRR(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRAs  t MinMaxNormcB@s5eZdZddddd„Zd„Zd„ZRS(sMinMaxNorm weight constraint. Constrains the weights incident to each hidden unit to have the norm between a lower bound and an upper bound. # Arguments min_value: the minimum norm for the incoming weights. max_value: the maximum norm for the incoming weights. rate: rate for enforcing the constraint: weights will be rescaled to yield `(1 - rate) * norm + rate * norm.clip(min_value, max_value)`. Effectively, this means that rate=1.0 stands for strict enforcement of the constraint, while rate<1.0 means that weights will be rescaled at each step to slowly move towards a value inside the desired interval. axis: integer, axis along which to calculate weight norms. For instance, in a `Dense` layer the weight matrix has shape `(input_dim, output_dim)`, set `axis` to `0` to constrain each weight vector of length `(input_dim,)`. In a `Convolution2D` layer with `dim_ordering="tf"`, the weight tensor has shape `(rows, cols, input_depth, output_depth)`, set `axis` to `[0, 1, 2]` to constrain the weights of each filter tensor of size `(rows, cols, input_depth)`. ggπ?icC@s(||_||_||_||_dS(N(t min_valueR trateR (RRR R R ((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR{s   cC@s}tjtjtj|ƒd|jdtƒƒ}|jtj||j|j ƒd|j|}||tj ƒ|9}|S(NR Ri( RRRRR RR RRR R(RRRR((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRs 0cC@s,i|jd6|jd6|jd6|jd6S(NRR R R (RR R R (R((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyRˆs   (R R RRRR(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyR^s cC@s t|ƒS(N(R(t constraint((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyt serializescC@st|dtƒd|ddƒS(Ntmodule_objectstcustom_objectstprintable_module_nameR!(Rtglobals(tconfigR$((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyt deserialize‘s  cC@s‚|dkrdSt|tƒr)t|ƒSt|tjƒr_it|ƒd6id6}t|ƒSt|ƒro|Std|ƒ‚dS(Nt class_nameR's*Could not interpret constraint identifier:( tNonet isinstancetdictR(tsixt string_typeststrtcallablet ValueError(t identifierR'((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pytget¨s    (t __future__RR-tRRtutils.generic_utilsRRtobjectRR RRRtmax_normtnon_negt unit_normt min_max_normtmaxnormtnonnegtunitnormR"R*R(R3(((s0/tmp/pip-build-isqEY4/keras/keras/constraints.pyts&  ' 3