RADAR.time_series.algorithms.modelsTransformersTS.autoformer

Submodules

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.attn module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.attn.AutoCorrelation(*args: Any, **kwargs: Any)[source]

Bases: Module

AutoCorrelation Mechanism with the following two phases: (1) period-based dependencies discovery (2) time delay aggregation This block can replace the self-attention family mechanism seamlessly.

forward(queries, keys, values, attn_mask)[source]
time_delay_agg_full(values, corr)[source]

Standard version of Autocorrelation

time_delay_agg_inference(values, corr)[source]

SpeedUp version of Autocorrelation (a batch-normalization style design) This is for the inference phase.

time_delay_agg_training(values, corr)[source]

SpeedUp version of Autocorrelation (a batch-normalization style design) This is for the training phase.

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.attn.AutoCorrelationLayer(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(queries, keys, values, attn_mask)[source]

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.decoder module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.decoder.Decoder(*args: Any, **kwargs: Any)[source]

Bases: Module

Autoformer encoder

forward(x, cross, x_mask=None, cross_mask=None, trend=None)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.decoder.DecoderLayer(*args: Any, **kwargs: Any)[source]

Bases: Module

Autoformer decoder layer with the progressive decomposition architecture

forward(x, cross, x_mask=None, cross_mask=None)[source]

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.DataEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.FixedEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.PositionalEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.TemporalEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.TimeFeatureEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.TokenEmbedding(*args: Any, **kwargs: Any)[source]

Bases: Module

forward(x)[source]
RADAR.time_series.algorithms.modelsTransformersTS.autoformer.embed.compared_version(ver1, ver2)[source]

:param ver1 :param ver2 :return: ver1< = >ver2 False/True

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.encoder module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.encoder.Encoder(*args: Any, **kwargs: Any)[source]

Bases: Module

Autoformer encoder

forward(x, attn_mask=None)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.encoder.EncoderLayer(*args: Any, **kwargs: Any)[source]

Bases: Module

Autoformer encoder layer with the progressive decomposition architecture

forward(x, attn_mask=None)[source]

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.endec module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.endec.moving_avg(*args: Any, **kwargs: Any)[source]

Bases: Module

Moving average block to highlight the trend of time series

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.endec.my_Layernorm(*args: Any, **kwargs: Any)[source]

Bases: Module

Special designed layernorm for the seasonal part

forward(x)[source]
class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.endec.series_decomp(*args: Any, **kwargs: Any)[source]

Bases: Module

Series decomposition block

forward(x)[source]

RADAR.time_series.algorithms.modelsTransformersTS.autoformer.model module

class RADAR.time_series.algorithms.modelsTransformersTS.autoformer.model.Autoformer(*args: Any, **kwargs: Any)[source]

Bases: Module

Autoformer is the first method to achieve the series-wise connection, with inherent O(LlogL) complexity.

forward(x_enc, x_dec, enc_self_mask=None, dec_self_mask=None, dec_enc_mask=None)[source]

Module contents