Recurrent

RNN

layer eddl::RNN(layer parent, int units, string activation = "tanh", bool use_bias = true, bool bidirectional = false, string name = "")

Fully-connected RNN where the output is to be fed back to input.

Parameters

• parent – Parent layer

• units – Dimensionality of the output space

• activation – Name of the activation function

• use_bias – Whether the layer uses a bias vector

• bidirectional – Wether the RNN is bidirectional or not.

• name – A name for the operation

Returns

The RNN layer

Example:

l = RNN(l, 32);

You can check a full example on Advanced models.

GRU

Gated Recurrent Unit - Cho et al. 2014.

layer eddl::GRU(layer parent, int units, bool mask_zeros = false, bool bidirectional = false, string name = "")

Gated Recurrent Unit (GRU).

Parameters

• parent – Parent layer

• units – Dimensionality of the output space

• mask_zeros –

• bidirectional – Wether the RNN is bidirectional or not

• name – A name for the operation

Returns

The GRU layer

layer eddl::GRU(vector<layer> parent, int units, bool mask_zeros, bool bidirectional, string name)

Example:

l = GRU(l, 128);

LSTM

layer eddl::LSTM(layer parent, int units, bool mask_zeros = false, bool bidirectional = false, string name = "")

Long Short-Term Memory layer - Hochreiter 1997.

Parameters

• parent – Parent layer

• units – Dimensionality of the output space

• mask_zeros –

• bidirectional – Wether the RNN is bidirectional or not

• name – A name for the operation

Returns

The LSTM layer

layer eddl::LSTM(vector<layer> parent, int units, bool mask_zeros, bool bidirectional, string name)

Example:

l = LSTM(l,32);

You can check a full example on Advanced models.