autopycoin.models.Stack

class autopycoin.models.Stack(*args, **kwargs)

A stack is a series of blocks where each block produces two outputs, the forecast and the backcast.

Inside a stack all forecasts are sum up and compose the stack output. In the meantime, the backcast is given to the following block.

Parameters

blockstuple[autopycoin.models.BaseBlock]

Blocks layers. they can be generic, seasonal or trend ones. You can also define your own block by subclassing BaseBlock.

Notes

input shape: N-D tensor with shape: (…, batch_size, time step). The most common situation would be a 2D input with shape (batch_size, time step).

output shape: N-D tensor with shape: (…, batch_size, units). For instance, for a 2D input with shape (batch_size, units), the output would have shape (batch_size, units). With a QuantileLossError with 2 quantiles or higher the output would have shape (quantiles, batch_size, units). If you add 2 variables, the output would have shape (variables, quantiles, batch_size, units).

Examples

>>> from autopycoin.layers import TrendBlock, SeasonalityBlock
>>> from autopycoin.models import Stack, NBEATS
>>> from autopycoin.losses import QuantileLossError
...
>>> trend_block = TrendBlock(label_width=20,
...                          p_degree=2,
...                          n_neurons=16,
...                          drop_rate=0.1,
...                          name="trend_block")
...
>>> seasonality_block = SeasonalityBlock(label_width=20,
...                                      forecast_periods=[10],
...                                      backcast_periods=[20],
...                                      forecast_fourier_order=[10],
...                                      backcast_fourier_order=[20],
...                                      n_neurons=15,
...                                      drop_rate=0.1,
...                                      name="seasonality_block")
...
... # blocks creation
>>> trend_blocks = [trend_block for _ in range(3)]
>>> seasonality_blocks = [seasonality_block for _ in range(3)]
...
... # Stacks creation
>>> trend_stacks = Stack(trend_blocks, name="trend_stack")
>>> seasonality_stacks = Stack(seasonality_blocks, name="seasonality_stack")
...
... # model definition and compiling
>>> model = NBEATS([trend_stacks, seasonality_stacks], name="interpretable_NBEATS")
>>> model.compile(loss=QuantileLossError(quantiles=[0.5]))

Attributes

blockstuple[autopycoin.models.BaseBlock]

Return the list of blocks.

label_widthint

Return the label width.

input_widthint

Return the input width.

is_interpretablebool

Return True if the stack is interpretable.

stack_typestr

Return the type of the stack.

property blocks: List[autopycoin.layers.nbeats_layers.BaseBlock]

Return the list of blocks.

property input_width: int

Return the input width.

property is_interpretable: bool

Return True if the stack is interpretable.

property label_width: int

Return the label width.

property stack_type: str

Return the type of the stack. CustomStack if the blocks are all differents.