Author Topic: Classify Images Decision Tree  (Read 3485 times)

robert gagliano

  • Sr. Member
  • ****
  • Posts: 422
    • View Profile
Classify Images Decision Tree
« on: June 10, 2011, 12:12:11 am »

1) Q: Is there a candidate
centered in the cross hairs of
the right-hand image?
No (-1) Yes

2) Q: Has the candidate itself
subtracted correctly?
No (-1) Yes

3) Q: Is the candidate star-like
and approximately circular?
No (-1) Yes (+1)

4) Q: Is the candidate centered in
a circular host galaxy?
No (+2) Yes

If answer to 2) is No then Q: What is wrong with the
subtraction?
No candidate
Not all pixels positive
Poor subtraction

If answer to 3) is No then Q: What is wrong with the
candidate?
Not circular - too small
Not circular - elongated
Not circular - distorted
Not circular - diffuse


The structure of the decision and scoring of the questions
means that candidates can only end up with a score of -1,
1 or 3 from each classification, with the most promising SN
candidates scored 3. As each new classification is received,
the arithmetic average score (Save) of the candidate is
recalculated. Candidates which are not astrophysically
interesting tend to have Save < 0 (i.e., most volunteers
scored them a ‘−1’). Astrophysical transients typically
have Save > 0, and SNE tend to have Save > 1 (i.e., most
volunteers scored them a ‘3’).

 Superzooites classified and scored 5% of the total candidates as good (3), 23% as average (1-3) and 72% as poor (-1).
 The "good" category had 60% and the "average" category 33% of the spectroscopically confirmed SNE discoveries. This data comes from the Galaxy Zoo Supernovae paper http://arxiv.org/abs/1011.2199

graham d

  • Hero Member
  • *****
  • Posts: 2408
    • View Profile
Re: Classify Images Decision Tree
« Reply #1 on: June 10, 2011, 10:26:17 am »
Congratulations to all of us - that it's a great success story is a bottom line summary.
 
I would like to make some technical comments on the classification system but I will not . The bottom line is it finds many sn's fast in time for spectral follow ups.

There's another of those remarkable  laws that if you study great tomes on  chess or poker strategy and tactics for instance, then your performance tumbles or is crippled for six months. Hence, I dare not make any comments lest I screw up my own classification strategy; let alone yours :). Years ago I used to classify tens of thousands of rock samples from time to time inorder to keep my eye in and limit advice to others on the front line. Don't overly interfere.
The only point I will comment on to the commanders is from 3.53 " My Supernovae" on page 6 by the 23 authors, quote
 - http://arxiv.org/PS_cache/arxiv/pdf/1011/1011.2199v2.pdf

Quote
3.5.3 `My Supernovae'
Providing feedback to the Galaxy Zoo Supernovae com-
munity is a vital part of the overall website experience to
encourage volunteers to return to the website. This is partly
done using forums and blogs where scientists can comment
on individual events classi ed by the zoo. In addition, each
volunteer can view a history of the candidates that they
have classi ed on their `My Supernovae' page.
The `My Supernovae' (MySN) page displays the can-
didate triplets. Those which have been observed are
overlaid with a small symbol identifying the candidate as
Galaxy Zoo Supernovae 7
a SN, variable star, or asteroid. Clicking on one of the
candidates also allows the volunteer to see the average
rating across all classi cations, the number of classi ers
and whether the candidate was selected for followup by
the PTF team. PTF observers are encouraged to leave
comments on the science dashboard that the classi ers can
also see on their MySN page.
« Last Edit: June 10, 2011, 10:28:57 am by graham d »

robert gagliano

  • Sr. Member
  • ****
  • Posts: 422
    • View Profile
Re: Classify Images Decision Tree
« Reply #2 on: June 10, 2011, 12:39:20 pm »
"My Supernova (MySN)" is no longer part of the Galaxy Zoo Supernovae homepage and was replaced by "Discoveries" when the website was updated. "MySN" reffered to in the paper had many glitches and became non functional for many zooites. "Discoveries" is an improvement because it identifies your discoveries by PTF ID which allows us to 1) find a discovery if it is published in ATel and 2) search "Discoveries" for a specific PTF ID. This is fyi for those newer to the SN Zoo who may be unfamiliar with the origin of graham d's comments.

Blackprojects

  • Hero Member
  • *****
  • Posts: 9518
    • View Profile
Re: Classify Images Decision Tree
« Reply #3 on: June 11, 2012, 11:12:33 am »
So a Core of 20% of SN-ZOOITES Does most of the Work from reading the paper Linked at the top compares with other ZOO"S


WE HAVE BEEN BUSY BUNNIES!

Blackprojects

  • Hero Member
  • *****
  • Posts: 9518
    • View Profile
Re: Classify Images Decision Tree
« Reply #4 on: June 11, 2012, 11:50:53 am »
NOW for Something Completly Diferent!

Dynamic Bayesian Combination of Multiple Imperfect Classifiers

"Edwin Simpson, Stephen Roberts, Ioannis Psorakis, Arfon Smith
(Submitted on 8 Jun 2012)

Classifier combination methods need to make best use of the outputs of multiple, imperfect classifiers to enable higher accuracy classifications. In many situations, such as when human decisions need to be combined, the base decisions can vary enormously in reliability. A Bayesian approach to such uncertain combination allows us to infer the differences in performance between individuals and to incorporate any available prior knowledge about their abilities when training data is sparse. In this paper we explore Bayesian classifier combination, using the computationally efficient framework of variational Bayesian inference. We apply the approach to real data from a large citizen science project, Galaxy Zoo Supernovae, and show that our method far outperforms other established approaches to imperfect decision combination. We go on to analyse the putative community structure of the decision makers, based on their inferred decision making strategies, and show that natural groupings are formed. Finally we present a dynamic Bayesian classifier combination approach and investigate the changes in base classifier performance over time. "


http://arxiv.org/abs/1206.1831

graham d

  • Hero Member
  • *****
  • Posts: 2408
    • View Profile
Re: Classify Images Decision Tree
« Reply #5 on: June 13, 2012, 08:54:29 pm »
What a title!
As a noble international team of people from diverse backgrounds and uncommon language we help to discover new supernovae fast.  Few people persevere at this task. Let us set the science detail aside. None of us are perfect. Who amongst us would have recognised this distorted image would become a supernova? Some stars do become supernovae regardless of image imperfection.



Perhaps the statisticians or even biologists who deal with base Nature categorise those individuals of species as such. There's too much 1984ish newspeak here. "human agents" are infact people; cohorts of imperfect decision makers" that's us folks or to keep it simple, the people again who hold certain truths to be self-evident...with the right to alter or institute new.....

There is no acknowledgement, this international cooperation of SN zooites as such, other than to a project; a citizen science project. Chris is indeed acknowledged so perhaps he could suggest a few improvements of a social kind; a very short and simple one. Without the enthusiasm and endeavours of an international group of citizens we the people... .
« Last Edit: June 13, 2012, 08:58:45 pm by graham d »

zutopian

  • Hero Member
  • *****
  • Posts: 2852
  • REOPEN this forum!!!
    • View Profile
Re: Classify Images Decision Tree
« Reply #6 on: March 11, 2013, 04:25:45 pm »
Warning: this topic has not been posted in for at least 120 days. (SN Zoo is retired.)
Here is an Object of the Day, which is related to this topic.:
http://www.galaxyzooforum.org/index.php?topic=280647.msg628332#new
« Last Edit: March 11, 2013, 05:05:42 pm by zutopian »