Heiko's Log

From Vision Wiki
Jump to navigation Jump to search

Aug 29 2007 - fbk on initial wing extension stats

Anderson sub-group meeting

Pietro and Heiko present very preliminary analysis of wing extension causes and effects. Carried out on Kii-Chi's male-male neuropeptide sequences.

Main comments:

  • (Pietro) Perhaps the data show that fly B approaching fly A causes A to extend wing, and that the consequence of that is that fly B runs away. This is very preliminary because (a) few data, (b) there could be bugs in the code, (c) the data do not show who is approaching whom. So: please consider this purely as an initialy hypothesis.
  • (Anderson, others) Should analyze first female-male sequences (Anne has 100 movies ready).
  • (Anderson, Liming) We should distinguish between wing flicks and wing extension. A better definition of wing extension: wing angle between 60 and 90 degrees, duration of extension should be considerable.
  • Anne sent Heiko examples of different kind of wing actions (flicks, extensions, ...)
  • It may be interesting to compare female-male and male-male interactions (in the context of measuring wing extension).
  • (Pietro) These are preliminary data: the red lines (random time periods) should be flat and sometimes they show structure (why?). Also: the green lines appear suspiciously different in the left-wing and right-wing extension cases.
  • (Anderson, Liming) One needs to be careful in drawing any cause-effect conclusions.
  • (Anne, Suzie) We have also observed that in male-male interactions fly B will `escape' after A does wing extension.
  • (Pietro) We need to generate a ground truth for wing extension and measure our % detectect and % false alarms carefully. Right now we do not know how well this is performing...

Meeting 1/16/2007 David, Pietro, Liming, Heiko

We have 2 projects:

  1. Fast throughput screen of genotypes using known behaviors: aggression, male-male courtship, locomotion
  2. Discovery of new behaviors
    1. Using ground-truth date coming from the 1st project: can we re-discover the behaviors that discriminate between different genotypes?
    2. Using raw data where we do not know the answer: the peptide project is a good candidate

We need different data for the 2 projects:

  1. Collect 10 movies each for 400-600 genotypes (half mutants and half circuit manipulations).
  2. Collect ~100 movies each for a small number of genotypes

Heiko and Liming will optimize the lighting for the 5-arena setup. Can we get movies that are clear enough for Heiko's software to analyze? Does the 5-arena setup save analysis time? (It saves storage of course).

The peptide movies are going to be collected in parallel using the single-arena setup.

Meeting 11/27/2006 David, Pietro, Lihi, Heiko

  • Classify behaviors based on whether they occur in flies of different genotypes
  • Use the given complete data set of approx. 80-90 movies
  • Can you tell which clusters are "interesting"?
  • Can you tell which cluster a fly uses?
  • Naming clusters
  • "Explaining" 2-fly behavior as compositions of individual flies vs. genuinely social behaviors
  • Repeat with blind flies, smell-blind flies etc. to study social behavior
  • Different time-frames within the same scheme
  • Ultimate goal: Find a new behavior
  • Portioning of behavioral time in flies development
    • Feeding
    • Fighting
    • Fleeing
    • Mating
  • At some points move to richer measurements
  • Is there a syntax to the order in which the clusters are occur