User Tools

Site Tools


play

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
play [2018/11/13 15:50] liyongplay [2018/11/14 13:50] (current) liyong
Line 32: Line 32:
 1 1
 </code> </code>
-When the learner constructs a conjectured NBA, it will ask an equivalence query for the conjecture. +When the learner constructs a conjectured NBA, it will ask an equivalence query for the following conjecture.
-In this case, you can answer a counterexample, say (b, ab) since the output automaton is not the right conjecture. +
 <code> <code>
 Resolving equivalence query for hypothesis (#Q=4)...   Resolving equivalence query for hypothesis (#Q=4)...  
Line 58: Line 57:
   4 -> 0 [label=""];   4 -> 0 [label=""];
 } }
 +
 +</code>
 +You can get the following graphical view of above dot text on the webpage [[http://www.webgraphviz.com/|Webgraphviz]] or by the xdot tool.
 +
 +
 +{{:a1.png?200|}}
 +
 +The above output automaton is clearly not the right conjecture.
 +In this case, you can answer a counterexample, say b(ab)<sup>ω</sup> given by its decomposition (b, ab), which is accepted by the conjecture but is not in the target language, to the learner.
 +<code>
 0 0
 Now you have to input a counterexample for inequivalence. Now you have to input a counterexample for inequivalence.
Line 67: Line 76:
 You input a loop: ab You input a loop: ab
 </code> </code>
- 
-{{:a1.png?200|}} 
  
 The learner will ask several membership queries in order to analyze the received counterexample and to construct a new conjecture. The learner will ask several membership queries in order to analyze the received counterexample and to construct a new conjecture.
Line 97: Line 104:
 0 0
 </code> </code>
-This time we receive the right conjectured NBA and reply with positive answer to the learner, which means that the learner has completed its learning task.+This time we receive the following conjectured NBA.
 <code> <code>
 Resolving equivalence query for hypothesis (#Q=3)...   Resolving equivalence query for hypothesis (#Q=3)...  
Line 116: Line 123:
 } }
  
 +</code>
 +
 +{{:a2.png?200|}}
 +
 +Above automaton is the right conjecture accepting the language {a,b}<sup>*</sup>b<sup>ω</sup>.
 +Thus we can reply with a positive answer to the learner, which means that the learner has completed the learning task.
 +<code>
 1 1
 Congratulations! Learning completed... Congratulations! Learning completed...
 </code> </code>
  
-Following command allows the user to see more details about the storage of the membership queries when using the table-based L<sup>$</sup> algorithm to learn an NBA.+Following command allows the user to see more details about the learning procedure.
 <code> <code>
 java -jar ROLL.jar play -periodic -v 2 java -jar ROLL.jar play -periodic -v 2
 </code> </code>
- +Besides the output conjectures, ROLL also allows the user to see how the FDFA and its corresponding observation tables evolve during the learning procedure. 
- +Moreover, the log also shows the counterexample analysis information as described in paper [6]. 
-The complete log information is given as follows.+  
 +The complete log information of an execution of above command is given as follows.
 <code> <code>
 PLAYING,TABLE,PERIODIC,NONE,NBA,UNDER,verbose=2,bs=false,dot=false,inputfile=null,outputfile=null,outputA=null,outputB=null PLAYING,TABLE,PERIODIC,NONE,NBA,UNDER,verbose=2,bs=false,dot=false,inputfile=null,outputfile=null,outputA=null,outputB=null
  
-2018/11/13 15:45:10 [ info] ROLL for interactive play...+2018/11/13 16:11:13 [ info] ROLL for interactive play...
 Please input the number of letters ('a'-'z'):  Please input the number of letters ('a'-'z'): 
 2 2
Line 197: Line 212:
 ab ab
 You input a loop: ab You input a loop: ab
 +Current FDFA:
 //FFA-M:  //FFA-M: 
 digraph { digraph {
Line 221: Line 237:
 Is w-word (b, ab) in the unknown languge: 1/0 Is w-word (b, ab) in the unknown languge: 1/0
 0 0
 +DFA D_{u$v} for counterexample (u, v):
 digraph Automaton { digraph Automaton {
   rankdir = LR;   rankdir = LR;
Line 245: Line 262:
 } }
  
-Not in target: b$ab +Counterexample not in target: b$ab 
-final counterexample b$ab+final counterexample for the FDFA learner: b$ab
 normalized factorization: (b, ab) normalized factorization: (b, ab)
 Is w-word (ϵ, ab) in the unknown languge: 1/0 Is w-word (ϵ, ab) in the unknown languge: 1/0
play.1542095435.txt.gz · Last modified: 2018/11/13 15:50 by liyong