From b8089d68837587c6262b5f7f0f919cad07dca78b Mon Sep 17 00:00:00 2001
From: lucas_miranda <lucasmiranda42@gmail.com>
Date: Wed, 19 May 2021 01:49:03 +0200
Subject: [PATCH] Replaced for loop with vectorised mapping on ClusterOverlap
regularization layer
---
deepof/model_utils.py | 32 +++++++++++++++++++-------------
deepof/models.py | 13 ++++++-------
2 files changed, 25 insertions(+), 20 deletions(-)
diff --git a/deepof/model_utils.py b/deepof/model_utils.py
index 064d4cb2..585623e9 100644
--- a/deepof/model_utils.py
+++ b/deepof/model_utils.py
@@ -612,19 +612,25 @@ class ClusterOverlap(Layer):
)
### CANDIDATE FOR REMOVAL. EXPLORE HOW USEFUL THIS REALLY IS ###
- neighbourhood_entropy = purity_vector # * max_groups[random_idxs]
-
- # self.add_metric(
- # len(set(hard_groups[max_groups >= self.min_confidence])),
- # aggregation="mean",
- # name="number_of_populated_clusters",
- # )
- #
- # self.add_metric(
- # max_groups,
- # aggregation="mean",
- # name="average_confidence_in_selected_cluster",
- # )
+ neighbourhood_entropy = purity_vector * tf.gather(
+ max_groups, tf.constant(random_idxs)
+ )
+
+ self.add_metric(
+ tf.shape(
+ tf.unique(
+ tf.squeeze(tf.gather(hard_groups, tf.where(max_groups >= self.min_confidence)))
+ )
+ ),
+ aggregation="mean",
+ name="number_of_populated_clusters",
+ )
+
+ self.add_metric(
+ max_groups,
+ aggregation="mean",
+ name="average_confidence_in_selected_cluster",
+ )
self.add_metric(
neighbourhood_entropy, aggregation="mean", name="neighbourhood_entropy"
diff --git a/deepof/models.py b/deepof/models.py
index c8a61fc3..10a18709 100644
--- a/deepof/models.py
+++ b/deepof/models.py
@@ -473,13 +473,12 @@ class GMVAE:
# Dummy layer with no parameters, to retrieve the previous tensor
z = tf.keras.layers.Lambda(lambda t: t, name="latent_distribution")(z)
- if self.overlap_loss:
- z = deepof.model_utils.ClusterOverlap(
- self.batch_size,
- self.ENCODING,
- self.number_of_components,
- loss_weight=self.overlap_loss,
- )([z, z_cat])
+ z = deepof.model_utils.ClusterOverlap(
+ self.batch_size,
+ self.ENCODING,
+ self.number_of_components,
+ loss_weight=self.overlap_loss,
+ )([z, z_cat])
# Define and instantiate generator
g = Input(shape=self.ENCODING)
--
GitLab