From: Mart Lubbers <mart@martlubbers.net>
Date: Tue, 26 Jan 2016 11:09:44 +0000 (+0100)
Subject: added part about automata learinng
X-Git-Url: https://git.martlubbers.net/?a=commitdiff_plain;h=009acb97a81050895a1aeb616e5e7b24952ff981;p=tt1516.git

added part about automata learinng
---

diff --git a/recap_al.txt b/recap_al.txt
index e69de29..7553159 100644
--- a/recap_al.txt
+++ b/recap_al.txt
@@ -0,0 +1,44 @@
+Introduction Model Learning
+===============================================================================
+Model learning black box
+- Passive: Process mining
+
+- Active: Automata learning
+	- Teacher and learner
+	- Learner can ask
+		- membership queries: yes/no from teacher
+		- equivalence queries: yes/no+counterexample from teacher
+	- Can be used for fingerprinting
+
+Active Automata Learning
+===============================================================================
+- Mealy machine
+- Nerode:
+	two words are equivalent if for all continuations they map the same output
+Observation table
+Columns: suffix closed
+Rows: prefix closed
+
+
+HYPOTHESIS
+- Two parts:
+	- top part: states
+	- bottom part: extensions
+- Closedness:
+	Bottom part structure should also be present in upper part otherwise move
+	up.
+- Completeness:
+	Drawing should be input complete, otherwise add rows with extensions
+
+EQUIVALENCE QUERY
+Yes: successfully found model
+No: Counterexample:
+	word w should give b
+
+HANDLE CE
+- extract suffix that disproves nerode
+	- there is a suffix that ends you up in the same state, the later part is
+	  the culprit.
+- add to column(note that columns should be suffix closed)
+- fill in table
+- refine hypothesis