Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • Jun 03 2019 21:43
    Uiuran commented #15
  • May 28 2019 12:58
    Uiuran commented #15
  • Mar 05 2017 06:18

    theideasmith on master

    updates to graph drawing algori… (compare)

  • Mar 03 2017 22:12

    theideasmith on master

    beginning study of neural field… Merge branch 'master' of https:… (compare)

  • Dec 19 2016 18:18

    slarson on master

    importing experiment to graph N… (compare)

  • Sep 19 2016 16:30

    slarson on master

    Create .openworm.yml (compare)

  • Jun 15 2016 12:06
    theideasmith commented #17
  • Jun 14 2016 18:40
    lukeczapla commented #17
  • Jun 14 2016 16:15
    theideasmith opened #17
  • Jun 13 2016 15:36
    theideasmith labeled #16
  • Jun 13 2016 15:36
    BenjiJack opened #16
  • Jun 05 2016 18:17
    slarson commented #14
  • Jun 05 2016 18:16
    slarson commented #14
  • Jun 05 2016 13:52
    theideasmith commented #14
  • Jun 03 2016 15:11
    theideasmith commented #14
  • Jun 03 2016 03:15
    lukeczapla commented #15
  • Jun 02 2016 16:08
    theideasmith commented #15
  • Jun 02 2016 16:05
    theideasmith commented #15
  • Jun 02 2016 02:20
    theideasmith commented #15
  • May 25 2016 17:53
    theideasmith edited #15
Stephen Larson
@slarson
@theideasmith These are great and it is excellent that you are playing in this space. It is very open ended, for sure! For the larger goal of getting a comparison working, my perspective is that if some of these analysis methods are yielding something that you could write a script that would 1) evolve a nervous system model in time 2) analyse its dynamics and 3) calculate a “score” for how similar or different those dynamics were from the Kato data; this would be hugely helpful. The key is to make a good argument for why the “score” is valid as a comparison. Then consider — if we had a very high-scoring network model; we could ask a lot deeper questions of the simulation along the lines you are interested in. We could ask very deep “why” questions about how the dynamics evolve in time, because we’d have all the variables and could work backwards in time from various dynamical events to understand in more depth why they happened. For me, that’s the complex systems approach to this question that will yield much more interesting hypotheses about neuronal dynamics.
Akiva Lipshitz
@theideasmith
Wow! Now I realize just how important these parameters are. So where should start with optimization methods. V1 can be very simple
I started describing a few things to look at in #7
Probably a first thing would be to see if c302NetTuner.py actually works currently, and then start some issues to get it into a working state if it doesn't
We can also use that as inspiration and depart from it and write something much simpler to get intuitions
Akiva Lipshitz
@theideasmith
A better optimization would compare simulated data to real data, but the kato datasets lack many neuron labells. I was thinking we could use machine learning classification and combinatorial optimization to label unlabelled neurons in the kato-datasets if a certain unlabelled time series has a high probability of being a labelled neuron in another dataset.
Stephen Larson
@slarson
It may be possible to narrow down which neurons the unlabelled ones are if we ask the right questions about the prep
:)
Chee Wai Lee
@cheelee
Hi I have a quick question on those 5 datasets ... are there any worm behavior context associated with each of them over the 18 minutes of time? They look like they are different, but I have also noticed that the order of the neurons-specific data along the Y-axis are different across the datasets. It might help if they had the same presented order, so either human eyes can pick out more patterns, or some sort of image differential analysis could key in on areas of difference between each of those datasets.
Akiva Lipshitz
@theideasmith
Yea - as of now we don't have behavioral data. We could ask L
Saul Kato for behavioral data used in his paper
Additionally, the plots you see have clustered the neurons, so they cluster differently in the different datasets. I've also got code to generate uniformly ordered plots. For those plots, I was just wondering how the clusters compare across the datasets. And welcome to neuronal analysis! Excited people are interested.
Akiva Lipshitz
@theideasmith
shared_datasets.jpg
If you'd like, here are the datasets plotted with uniform ordering; you do lose clustering, but gain cross-dataset visual comparability
shared_datasets.jpg
Akiva Lipshitz
@theideasmith
Inspired by DevoWorm, I've made a project waffle.io board: https://waffle.io/openworm/neuronal-analysis
Akiva Lipshitz
@theideasmith
/all I think it would be productive to have a project meeting sometime. When are people available?
Stephen Larson
@slarson
Yeah let's do it! Maybe next Friday?
Akiva Lipshitz
@theideasmith
That would work for me.
Stephen Larson
@slarson
@theideasmith hey actually are you around now?
Akiva Lipshitz
@theideasmith
Yeah
Stephen Larson
@slarson
OK — hangout in like 10 minutes?
Akiva Lipshitz
@theideasmith
My previous message didn't go through. I'm not able to talk today. Monday?
Stephen Larson
@slarson
Ah sorry about that! No worries. Monday is tight for me; Friday is safest to plan for now but maybe we can do something ad hoc on Monday
Stephen Larson
@slarson
@/all sent an invite for Friday at 2pm ET for a chance to sync. If others want to join, just message me here in the channel
By the way, found this paper and it seems really relevant: http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1000190
Akiva Lipshitz
@theideasmith
@slarson can you resend the invite?
Stephen Larson
@slarson
Yes; doing now
Stephen Larson
@slarson
Thanks for making that waffle board @theideasmith. Was just updating it.
Akiva Lipshitz
@theideasmith
timeseries0.jpg
Initial simulations of model used in Organization of Excitable Dynamics in Hierarchical Biological Networks
Luke Czapla
@lukeczapla
It's pretty interesting results and I just made a minor change to align the model exactly with the paper. I haven't had luck with PyOpenWorm so the CElegansNeuronTable.xls was used here. I think tuning simple models with experimental data is a good plan to demonstrate the proof of principle but this one is too simplistic. If individual neurons at least had a few parameters (individual f and p values is a good start - it'd give a coarse-grained picture of neuronal excitability and recovery) then we could extract information by fitting to the kato data and do a full scientific simulation study
Stephen Larson
@slarson
@theideasmith @lukeczapla Wow :) OK we should find a time to get together and discuss these. This week I’m all booked up; how’s your next week looking?
Akiva Lipshitz
@theideasmith
Any day next week
Akiva Lipshitz
@theideasmith
The model I used to generate the figure above was written incorrectly. Here's what it really should look like:
timeseries.jpg
Akiva Lipshitz
@theideasmith
Three State Model on Multiple Graph Types, with same number of nodes
Multiple_Graphs_Sim.jpg
Stephen Larson
@slarson
I totally want to know more!
Ok @lukeczapla any constraints on day next week?
Akiva Lipshitz
@theideasmith
/all any time to meet tomorrow or Friday?
Stephen Larson
@slarson
Hey hey you should have an invite from me already
Maybe i sent it to the wrong address?
It is for friday 3pm ET
Akiva Lipshitz
@theideasmith
Ok great
Stephen Larson
@slarson
@theideasmith @lukeczapla ready when you are
Stephen Larson
@slarson
@theideasmith @lukeczapla Sorry we couldn’t connect today
Luke and I were chatting a little on here
He was saying: