From 882f60d2ed9057e05b4b73de12ccd32bfc02e6d2 Mon Sep 17 00:00:00 2001
From: Peter Fackeldey <peter.fackeldey@rwth-aachen.de>
Date: Fri, 19 Feb 2021 16:26:47 +0100
Subject: [PATCH] add permutation inveriant dense prototype layer
---
keras.py | 56 +++++++++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 55 insertions(+), 1 deletion(-)
diff --git a/keras.py b/keras.py
index 614688c..3625349 100644
--- a/keras.py
+++ b/keras.py
@@ -450,7 +450,6 @@ class PlotMulticlass(PlottingCallback, TFSummaryCallback):
def on_train_end(self, logs=None):
self.make_eval_plots()
-
def make_input_plots(self):
inps = self.x
if not isinstance(inps, (list, tuple)):
@@ -1333,6 +1332,61 @@ class Xception1D(tf.keras.layers.Layer):
return {"sub_kwargs": self.sub_kwargs}
+class PermutationInvariantDense1D(tf.keras.layers.Layer):
+ """
+ The PermutationInvariantDense1D layer is a TimeDistributed(Dense) implementation.
+ Additionally it imposes (if wanted) a permutation inivariant pooling operation.
+
+ Parameters
+ ----------
+ name :
+ name of the layer
+ depth :
+ depth of the ResNet layers
+ nfeatures:
+ number of features of Conv1D layer
+ batchnorm:
+ enable batchnorm for ResNet layers
+ pooling_op:
+ permutation invariant pooling operation
+ """
+
+ def __init__(self, **kwargs):
+ name = kwargs.pop("name", "PermutationInvariantDense1D")
+ super().__init__(name=name)
+ self.depth = kwargs.pop("depth", 1)
+ self.nfeatures = kwargs.pop("nfeatures", 32)
+ self.batchnorm = kwargs.pop("batchnorm", False)
+ self.pooling_op = kwargs.pop("pooling_op", None)
+
+ def build(self, input_shape):
+ opts = dict(activation="elu", kernel_initializer="he_normal")
+ self.layer = tf.keras.layers.TimeDistributed(
+ tf.keras.layers.Dense(
+ self.nfeatures,
+ input_shape=input_shape,
+ **opts,
+ )
+ )
+ network_layers = []
+ for layer in range(self.depth - 1):
+ network_layers.append(
+ tf.keras.layers.TimeDistributed(tf.keras.layers.Dense(self.nfeatures, **opts))
+ )
+ if self.batchnorm:
+ network_layers.append(tf.keras.layers.BatchNormalization())
+ if self.pooling_op:
+ # permutation invariant pooling op
+ network_layers.append(tf.keras.layers.Lambda(lambda x: self.pooling_op(x, axis=1)))
+ self.network_layers = network_layers
+
+ def call(self, input_tensor, training=False):
+ x = input_tensor
+ for layer in self.network_layers:
+ x = layer(x, training=training)
+ return x
+
+
class SplitHighLow(tf.keras.layers.Layer):
def call(self, inputs):
return inputs[:, :, :4], inputs[:, :, 4:]
--
GitLab