Tuesday, March 27, 2012

Geo-centric model and limit of Occam's razor.

I mentioned geo-centric model of the planetary motion a couple of times about how a sufficiently complex model can always be invented to fit the data.

In case you haven't seen a geo-centric model before, here is a Ptolemaic one

(see wikipedia entry for Occam's razor--another term I used in class about Nature's predilection for simplicity--which interestingly also uses helio-centric and geo-centric models as examples:  http://en.wikipedia.org/wiki/Occam's_razor


ps: interestingly, nature doesn't always seem to go for the "simplest" model--or at least not necessarily as we will model "simple". An interesting case in point is how the genes code the 20 amino acids. The original idea Crick had involved coming up with exactly 20 "unambiguous" triplets that can be placed next to each other and you will still be able to decode the sequence correctly even though there is no obvious delimiters between each amino acid code. This "comma free" version turned out not to be the one nature uses--which uses ambiguous triplets but ensures correct translation by using  specific starting/stopping  delimiters (see for example  http://www.emunix.emich.edu/~rwinning/genetics/code.htm).  The original Crick code was so compelling that it has been called the "greatest wrong theory". 

Re: Project 3 bn/bif question

I think .bif is fine (the .bn was the format they used to support in the past...).


On Tue, Mar 27, 2012 at 10:16 AM, Ivan Zhou <izhou@asu.edu> wrote:
Quick question: In the project when it is asking for .bn, is it refering to the .bif text since that is the first option under the edit toolbar command? Or is it just copying only the probabilities section that the .bif text shows at the bottom? I'm confused since I cannot seem to find any "bn" reference in the applet.

Ivan Zhou
Graduate Student
Graduate Professional Student Association (GPSA) Assembly Member
Eta Kappa Nu (HKN) Active Member 
School of Computing, Informatics and Decision Systems Engineering
Ira A. Fulton School of Engineering
Arizona State University

Monday, March 26, 2012

Project 3 submission instructions..

I received some questions on project 3 submission format. 

As mentioned in the project description, all we need from you is a report (in hard copy) showing your answers/observations to all the tasks.
In all cases, include the copy of the bayes network (and CPTs) that the applet showed.


Sunday, March 18, 2012

Re: Dawkins' rant on coin-toss superstitions

yes that is interesting reading ;-) 

On Sun, Mar 18, 2012 at 4:09 PM, Srinath Raghavan <sraghav4@asu.edu> wrote:
Hello Professor,

I read this after the last class. What a coincidence! Couldn't resist sharing it with you.

Page 235/271
Prof Richard Dawkins' rant on coin tossing superstition  http://i.imgur.com/zyWfv.jpg 


Saturday, March 17, 2012

textbook writeup on D-Sep condition

As some of you pointed out, the latest edition of R&N doesn't have discussion on the D-Sep condition--which I
did in the class. 

You can get the discussion from an earlier edition at http://rakaposhi.eas.asu.edu/cse471/dsep-condition.pdf 


Homework 2 will be due on 3/29

As discussed in the class, 
Homework 2, on topics of planning graph, propositional logic and bayes networks, will be due on Thursday after springbreak.


Tuesday, March 13, 2012

Grading for your project 1

Hi all:

You will receive the grade for your project 1 today. Here is the stats:
- Undergrad: max = 120, min = 30, avg = 98.7, stdev = 27.4
- Graduate: max = 126 (including 5% bonus), min = 55, avg = 107.3, stdev = 17.8

The detail of each part is as follows:

PART 1: Total = 40
Task 1: 5, task 2: 15 (for the right, up and down moves), task 3: 5, task 4: 10 (5 for each heuristic function), and task 5: 5.
If your code produced the correct sequence of moves for the 5 instances, then you will very likely get the full credits. One exception I saw was that some student implemented the heuristics incorrectly, and thus the statistics and the corresponding analysis did not make sense.

PART 2: Total = 38 
Task 1A, B, C, E: 5 for each, task 1D: 8. Task 2: 10.
This is perhaps the easiest part of the code, though some students got the branching factors wrong.

PART 3: Total = 42
For each item listed in part 2, you need to have the statistical numbers and reasonable comments on the difference between (a) the blind search and the informed search algorithms, and (b) contrasting the A* search with the two heuristics. The full credit for each one is as follows: 1A: 10, 1B: 6, 1CE: 6, 1D: 15 and 5 for the running time. 
Initially, I expected to see the statistics when running your code with a large number of test cases (not only the one provided)---large enough to see especially the clear distinction between the two heuristics; however, the reality is that very few had additional test cases. So my "relaxation" for this part is that you will get 1/3 credits for each of the followings: (a) correctly reported the numbers for the test cases provided, (b) correctly described the difference between the breath-first and the informed search approaches based on your data, and (c) similar to (b) but between the Manhattan and the misplaced tiles heuristics. 

Please let me know if you have any question.

Friday, March 9, 2012

Project 3 released; Due 3/27


 I released the next project--this will involve modeling with Bayes networks (and using a bayes net applet).  It will be due the Tuesday after the Spring Break. 

This project is also modeling rather than a coding project. It dovetails well with the bayes network homework question (you can use the applet to check your homework question answers, for example).

I may add another small coding part to the project before the due date.


Thursday, March 8, 2012

Monday, March 5, 2012

Office hours tomorrow from 2:30--3:30pm (instead of 1-2pm)


 I have to shift my office hours to 2:30pm tomorrow.


Project 2: Extra Credit


We've had a few questions about possible extra credit for Project 2. Here's a clarification that may (or may not) help you.

Loosely, what we are looking at is for you to explore the boundaries of what you can and cannot model with PDDL. Given that, something like just throwing a larger number of objects into a problem simply to make a bigger instance is probably not very interesting. 

However, exploring the things you can do (for example) using the type hierarchies in the definition of the object types, defining different actions from the ones mentioned, extending the domain itself - these are all interesting things to do. These will of course lead to problem instances different from the ones that we have provided. Of course, these are only some suggestions.

Please do include a detailed explanation of whatever extra credit you choose to do (if at all) in your project report. We will try our best to provide you support if you run into problems while pursuing the extra credit, but in the interests of fairness we will try to help students who have questions about the mandatory portion of the project first.

Good luck.

Project 2: Submission Instructions

Hello -

Here are some instructions on submitting project 2. We will be using Blackboard for the submision (like Project 1). There should be a link to submit it under course content on the Blackboard page. Please read through this entire mail.

Please include all your deliverables in one zipped folder, which you will name proj2_lastname_firstname.zip (so if your name is 'John Smith', your zip file should be proj2_smith_john.zip). The list of deliverables is available on the specification webpage.

Please make sure that you only submit once, as this makes it much easier for us all. The due date is Thursday March 8th 2012 before class (so any time before 4:30pm Arizona time is fine).

Additionally, you will turn in a printed copy of your project report in class. If you find that you are unable to make it to class that day, please hand it to one of the TAs or leave it in Dr. Rao's mailbox before then.

Thanks and good luck.

Flip-flopping--from a cognitive, social and effectiveness stand points..

The last couple of classes, we had occasion to talk about "flip-flopping" as an analogy for the monotonicity of logical inference. 

This morning, I heard a very interesting news story on NPR about how people perceive inconsistency as well 
as whether consistent people are more effective than inconsistent ones.  Here is a link; you might enjoy it: