initial commit

hamadichihaoui · Oct 14, 2021 · 30b5fdf · 30b5fdf
1 parent 07a7c23
commit 30b5fdf
Show file tree

Hide file tree

Showing 7 changed files with 54 additions and 1 deletion.
diff --git a/.idea/.gitignore b/.idea/.gitignore
diff --git a/.idea/inspectionProfiles/profiles_settings.xml b/.idea/inspectionProfiles/profiles_settings.xml
diff --git a/.idea/modules.xml b/.idea/modules.xml
diff --git a/.idea/unsupervised_anomaly_detection.iml b/.idea/unsupervised_anomaly_detection.iml
diff --git a/.idea/vcs.xml b/.idea/vcs.xml
diff --git a/README.md b/README.md
@@ -1 +1,24 @@
-# unsupervised_anomaly_detection
+# unsupervised_anomaly_detection
+
+## Dataset Type
+
+Based on the used dataset in the training dataset
+- Contaminated dataset: it contains both normal and anamolous elements in the training. 
+- Uncontaminated dataset: The training data contains only normal samples. The methods that take this assumption are also refered weakly supervised approaches.
+
+## UFDGAN 
+#### Assumptions: 
+Contaminated training set (the anomaly percentage is known)
+
+
+#### Why based on the unsupervised contrastive learning will not work?
+
+The dataset is composed of multi modal data, but only one mode is labaled as anomaly. If the dataset is to a human, he will not be able to the anomaly exactly.
+For that, the  
+
+<img src="assets/training_samples.JPG" alt="drawing" width="50%" height="50%"/>
+
+
+
+
+## Mode collapse
diff --git a/assets/training_samples.JPG b/assets/training_samples.JPG