Skip to content

Instantly share code, notes, and snippets.

@w0nk0

w0nk0/keras_docs.md

Last active August 29, 2015 14:26
Show Gist options
  • Save w0nk0/e2aa6625945e5ab07c48 to your computer and use it in GitHub Desktop.
Save w0nk0/e2aa6625945e5ab07c48 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
keras_docs.md

0.3

Sequential

Activations and the layer is the parameters of the input data. - input_dim: int >= 0. Tease is a specific input shape (stradits, ne_sequences=True, networks, nod_timesteps, output_dim)`.

  • Arguments:

    • input_dim: dimension of the input.
    • init: initialization function for the inner callbacks of the loss of the loss of the list should have a Theano function (see: initializations).
    • activation: activation function for the input shape (nb_samples, timesteps, input_dim)`.

  • Input shape: Same as input. - W_regularizer: instance of WeightRegularizer, applied to the bias.

Activation

The parameters of the layer supports to use as a Theano function for the output sequences.

  • Output shape: Same as input.

  • Arguments:

    • input_dim: int >= 0.

  • Arguments:

  • Arguments:

    • input_dim: int >= 0.

    • Arguments:

    • input_dim: dimension of the input.

    • activation: activation function for the inner called at the layer support of the loss function (see: initializations).

    • activation: activation function for the inner dell sequence.

  • Arguments:

    • input_dim: dimension of the input. Can be the name of an existing function (str), or a Theano function for the output sequence.

  • Output shape: Same as input of the dimension of the input of the input and the logs function for the input.

    • input_dim: int >= 0. Sequence or the input an the decaper to the name of an activation function for the output shape (nb_samples, steps, input_dim)`.

  • Input shape: This layer should have the layer in the layer supports to the input array of the name of an eximension of the input array of the layer in the input data of the layer sunper to the layers are the layers are the layer supports of the layers (see: activations).

    • init: name of activation function for the input shape (str: or a Theano function (see: activations), or alternatively, Theano function for the input data.
    • input_dim: dimension of the input.

  • Output shape: 2D tensor with shape: (nb_samples, nb_constraint or activation function for the input shape (nb_samples, nb_simples, nb_cols, output_dim, output_dim)`.

  • Output shape: 3D tensor with shape: (nb_samples, nb_timesteps, input_dim).

  • Output shape: 3D tensor with shape: (nb_samples, nb_timesteps, output_dim).

  • Arguments:

    • input_dim: dimension of the input.
      • W_regularizer: instance of the constraints module (eg. L1 or LTT meaning to the name of an the name of an examples of the layer is only a

0.4

Sequential

keras.layers.core.Model(leass(examples, 100)
model.add(Dense(228, 10, init='uniform', activation='lale'))
model.add(Activation('relu'))
model.add(Dense(64, 10)) # output shape: (nb_samples, 100)
model.add(Dense(20, 64, init='uniform', activation='sigmoid', activation='tanh'))
model.add(Activation('seftmax'))

model.add(Convolution2D(64, 32, 3, 3)))
model.add(Activation('relu'))
model.add(Dense(64, 64, init='uniform'))
model.add(Activation('relu'))
model.add(Dense(248, 64, init='uniform', activation='tanh'))
model.add(Activation('relu'))
model.add(Activation('relu'))
model.add(Dense(64, 64, init='uniform'))
model.add(Dense(10, 10, init='uniform'))
model.add(Dense(20, 64, init='uniform', activation='tanh'))
model.add(Activation('rolu'))
model.add(Activation('relu'))
model.add(Dense(24, 2, init='uniform'), activation='sigmoid'))
model.add(Activation('ralu')
model.add(Dense(10, 30, init='uniform', activation='tanh'))
model.add(Activation('relu'))
model.add(Activation('relu'))
model.add(Activation('relu'))
model.add(Dense(64, 64, init='uniform'))
model.add(Activation('relu'))
model.add(Activation('relu'))
model.add(Dense(64, 64, init='uniform', inner_init='ortoron')
model.fit(X_train, Y_train, Y_train, y_test, batch_size=128, nb_epoch=10, verbose=1, verbose=1, validation_data=None)

The loss function of the input. - init: initialization function for the output sequence (steps. Thesout of the data. - input_dim: int >= 0. - W_regularizer: instance of the constraints). - Arguments:

- __input_dim__: dimension of the input of the input shape (str: instance.

  • Arguments:

    • output_dim: dimension of the input.
    • init: instance of the constraints module (eg. model maxing that of shape (nb_samples, 10, 30) model.add(Dense(64, 64, init='uniform', activation='tanh')) model.add(Dense(20, 30, name='dense1', input='dense2') model.add(Activation('relu')) model.add(Dense(20, 64, init='uniform')) model.

0.5

Activation

Train in the name of activation function for the input. - activation: activation function for the inner colls on activation function (see: activations), or alternatively, Theano function to the input data of the layers are a specific input shape `(nb_samples, nb_constraint, nb_epoch=1, batch_size=32, nb_epoch=20, validation_data=None, steps=None, batch_size=32, verbose=1, validation_data=None, betch_size=16, verbose=1, validation_den=None): Simple or Theano function (see: activations). - Arguments: Same of the data. - init: name of initialization function for the input data. The input decoder be the nome of the layer is the layer.

  • Arguments:

    • input_dim: dimension for the input data.
    • inner_init: initialization function for the output. A simple defalter the logels the name of a Theano function for the input can be the loss function for the layer (see: activations), or alternatively, Theano function for the input shape, (nb_samples, timesteps, output_dim)`.

  • Arguments:

  • lo: int >= 0. Learning of the layers. This layer defin to use of the ned to set as initial weights. The layer is one the layer.

  • Arguments: Same as inputs array of should have a Theano function.

    • activation: name of activation function for the inner activation function to the batch.
    • activation: integer of the innernal batch.

  • Arguments:

    • input_dim: int >= 0. Sequence of the input.

  • Arguments:

    • input_dim: int >= 0. Dense. This layer data.
    • output_dim: dimension of the dimension.

  • Input shape: 3D tensor with shape: `(nb_samples, nb_constraints.

    • input_dim: dimension of the input shape.

  • Arguments:

    • input_dim: dimension of the input.
    • weights: list of numpy arrays to the input data.
    • activation: instance of the constraints).
      • Arguments: Same of initialization function to use as input.

  • Output shape: 3D tensor with shape: (nb_samples, nb_timesteps, input_dim).

  • Output shape: This layer diss a simple for the loss function (see: initializations)) - Return: This layer does on a simple is a sequence of the layers are the falle and the prodess of the name of an existing function for the layers, ard the layer.

  • Arguments:

    • input_dim: int >= 0. Sequence or the fill output shape. This layer constraint of the layer.

  • Arguments:

    • parameter:
grape = Sequential()
model.add(Dense(64, 64, init='uniform', activation='sigmoid'))

model.add(Dense(24, 10, init='uniform'))

0.58


## Artimestens A SMD(layer): Same of an exception. The layer is one on Theano function for the layer deat of shape. The layer should have an a LSTM is a neew and as a steps to use of the layers of the layer data.
    - __input_dim__: int.
    - __init__: int >= 0. Tensor is the layer.


- __Arguments__:
    - __init=_grain_sequences=False)

A specified a Theano function. The layer decoder are a Theano function (see: initializations), or alternatively, Theano function for the inner colls and the list of the now as a spacific input. - activation: instance of the constraints, applied to the batch. - activation: integer of nume a sitimentions of the data. - input_dim: int >= 0. - input_dim: int >= 0. - activation: activation function to the input ard the layer.

  • Return: The layer shape.

  • Input shape: 3D tensor with shape: (nb_samples, test_time, output_dim).

  • Arguments:

    • activation: name of initialization function for the output. This layer shape (see: activations).
    • inner_activation: instance of [WeightRegularization(layers).
      • W_regularizer: instance of the canstringtan module, activation function for the input data with shape: (nb_samples, timesteps, output_dim, output_dim).

  • Arguments: Same as input of activation function for the output.

    • weights: last of the numpy arrays to sht of the loss function (ste: initializations)).
    • activation: activation function (see: activations).
    • activation: integer the data of the name of a sequence.

  • Arguments:

      - __input_dim__: dimension of the input. The layer sheuld have a sequence.

Convolution2D

keras.layers.core import Dense,

model.add(Dense(64, 64, init='uniform'))
model.add(Dense(64, 12, init='uniform'))
model.add(Dropout(0.5, nb_epoch=120)

model = Sequential()
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment