| --- Day changed Wed Sep 03 2008 | ||
| -!- elias` [n=me@unaffiliated/elias/x-342423] has quit [Read error: 145 (Connection timed out)] | 00:13 | |
| -!- elias` [n=me@unaffiliated/elias/x-342423] has joined #hplusroadmap | 00:22 | |
| ybit | kanzure, whenever you are through grabbing the vids from superkuh, want to torrent them? | 00:34 |
|---|---|---|
| -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has quit ["Leaving"] | 01:29 | |
| -!- nsh [n=nsh@eduroam-52.uta.fi] has joined #hplusroadmap | 02:16 | |
| -!- Depucelator [n=Depucela@c-24-118-116-62.hsd1.mn.comcast.net] has quit ["Leaving."] | 02:36 | |
| -!- Nofaris [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has quit [Read error: 110 (Connection timed out)] | 02:51 | |
| -!- jm|earth [n=jm@p57B9DD0F.dip.t-dialin.net] has joined #hplusroadmap | 03:31 | |
| -!- marainein [n=marainei@220.253-204-179.VIC.netspace.net.au] has joined #hplusroadmap | 04:39 | |
| kanzure | ybit: Yes. | 07:13 |
| kanzure | http://www.bbc.co.uk/radio4/redesigninghumanbody/pip/lrgnp/ | 07:15 |
| kanzure | http://www.bbc.co.uk/radio4/science/humanalien.shtml | 07:15 |
| procto | I can't wait to start ripping out body parts and replacing them with improved versions | 08:38 |
| procto | I don't use computer from the 80's | 08:38 |
| procto | why does my biology have to be from the 80's? | 08:38 |
| -!- Phreedom__ [n=freedom@ip-194-50-167-184.mir.dn.ua] has joined #hplusroadmap | 08:52 | |
| -!- Phreedom_ [n=freedom@ip-194-50-167-184.mir.dn.ua] has quit [Read error: 110 (Connection timed out)] | 08:53 | |
| -!- marainein [n=marainei@220.253-204-179.VIC.netspace.net.au] has quit [Read error: 113 (No route to host)] | 09:23 | |
| -!- ybit [n=h@unaffiliated/ybit] has quit [Connection timed out] | 09:46 | |
| -!- pk [i=pk@cash-3b-130.res.umass.edu] has joined #hplusroadmap | 11:08 | |
| kanzure | ideas on representing reaction kinetics ? | 12:12 |
| kanzure | 'stochastic kinetics modeling of biochemical components' | 12:12 |
| kanzure | is SBML or CellML something I should be looking into? | 12:12 |
| kanzure | http://sbml.org/Basic_Introduction_to_SBML | 12:12 |
| kanzure | http://sbml.org/SBML_Software_Guide/SBML_Software_Matrix | 12:13 |
| kanzure | one of those is a 'stochastic framework' | 12:13 |
| kanzure | BioNetGen http://bionetgen.org/ | 12:13 |
| kanzure | BioPathwise http://bioanalyticsgroup.com/ | 12:13 |
| kanzure | BioNetGen being the free one | 12:14 |
| kanzure | huh: http://www.yeastpheromonemodel.org/wiki/Main_Page | 12:14 |
| kanzure | okay, yeah | 12:14 |
| kanzure | http://bionetgen.org/index.php/Tutorial_Example | 12:14 |
| kanzure | reaction rules are in this document | 12:14 |
| kanzure | but this looks like a small project ... | 12:14 |
| kanzure | those parameters are very useful | 12:15 |
| kanzure | if I could map a database of biobricks to BNGL files. | 12:15 |
| kanzure | http://193.146.190.210:8080/ByoDyn ' Software package for the analysis of paratemer landscapes in biological networks' | 12:22 |
| kanzure | note: /mnt/maxtor/all/Biology/ has http://www.ebi.ac.uk/biomodels/release_21August2008_sbmls.tar.gz now | 12:25 |
| -!- Nofaris [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has joined #hplusroadmap | 12:46 | |
| -!- Phreedom_ [n=freedom@ip-194-50-167-184.mir.dn.ua] has joined #hplusroadmap | 13:12 | |
| -!- Phreedom__ [n=freedom@ip-194-50-167-184.mir.dn.ua] has quit [Read error: 110 (Connection timed out)] | 13:13 | |
| kanzure | http://heybryan.org/projects/cerebellum/ <-- this is a git repo (despite the lack of a .git ending to the name) | 13:13 |
| kanzure | has mostly visual basic .. I apologize | 13:13 |
| kanzure | am forking slowly. | 13:13 |
| kanzure | Haha, 'kinetic simulation and model fitting' is being presented at the old lab as of 15 minutes ago | 13:14 |
| kanzure | which is exactly what Dan and I were talking about | 13:14 |
| kanzure | where is faceface anyway? | 13:14 |
| -!- Nofaris_ [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has joined #hplusroadmap | 13:35 | |
| -!- Nofaris [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has quit [Read error: 110 (Connection timed out)] | 13:50 | |
| -!- Nofaris_ [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has quit [Read error: 110 (Connection timed out)] | 14:26 | |
| -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has joined #hplusroadmap | 15:58 | |
| kanzure | degradation rates, gene expression rates, polymerase per second (PoPS), what else might you want to 'optimize' in a GRN of biobricks? | 16:25 |
| -!- ybit [n=h@unaffiliated/ybit] has joined #hplusroadmap | 17:10 | |
| -!- biopunk [n=p@h91n3c1o261.bredband.skanova.com] has joined #hplusroadmap | 17:25 | |
| -!- jm|earth [n=jm@p57B9DD0F.dip.t-dialin.net] has quit ["quit"] | 17:31 | |
| kanzure | is anybody going to kill me if I just do basic stoichiometry? | 17:38 |
| kanzure | http://bionetgen.org/index.php/Tutorial_Example | 17:38 |
| kanzure | look at "begin reaction rules" | 17:38 |
| kanzure | I'm pretty sure that's just stoich | 17:38 |
| * bkero has to do stoiciometry on his car daily. :/ | 18:07 | |
| kanzure | buh? | 18:13 |
| kanzure | gas? | 18:14 |
| bkero | Yea | 18:14 |
| bkero | Burning too hot will make gaskets brittle and lose compression | 18:14 |
| kanzure | I should just make it variable for variables to optimize for | 18:15 |
| kanzure | Now I just have to worry about how to make the hidden variable model itself | 18:15 |
| kanzure | Say that I have 10,000 different designs | 18:15 |
| kanzure | and I can simulate all of them and tweak all of them | 18:15 |
| kanzure | now which way do you tweak and such? This is what an ANN would be for, supposedly | 18:16 |
| bkero | ANN? | 18:16 |
| kanzure | but do you just set the exact outputs of the ANN to the parameters for each of the designs ?? | 18:16 |
| kanzure | ANN or hidden markov model or something else | 18:16 |
| kanzure | different ways are possible there | 18:16 |
| bkero | Standardize some of the components of a design and inherit your variable as a class? | 18:18 |
| kanzure | What? | 18:19 |
| bkero | You're trying to find a uniform way to tweak different designs? | 18:22 |
| kanzure | yes, that sounds right. | 18:22 |
| bkero | In an object model, that's what object extension or inheritence is for. | 18:23 |
| kanzure | You mean I've been using object oriented programming incorrectly for the past, uh, decade? | 18:24 |
| bkero | What have you been doing 'wrong'? | 18:24 |
| kanzure | I usually just use hierarchical classes, so I have parents and siblings and so on. | 18:24 |
| kanzure | (class inheritance) | 18:25 |
| kanzure | and then object instantiation is treated just like constructing a new instance of a data struct. | 18:25 |
| bkero | Sure | 18:28 |
| bkero | But you can extend classes | 18:28 |
| kanzure | yeah? | 18:28 |
| kanzure | virtual class members, you mean? | 18:28 |
| bkero | java104/ | 18:29 |
| bkero | http://www.javacoffeebreak.com/java104/java104.html | 18:29 |
| bkero | Virtuals in java are like generics, right? | 18:29 |
| kanzure | Hm. | 18:30 |
| kanzure | I forget. | 18:30 |
| bkero | chizu: Does python have class extension ala java? | 18:31 |
| chizu | Yes, Python is a typical class based OO language. It works more or less how Java does. | 18:39 |
| -!- Nofaris_ [n=Nofaris@adsl-75-42-77-105.dsl.scrm01.sbcglobal.net] has joined #hplusroadmap | 18:39 | |
| kanzure | ' Fujaba uses Story driven modelling, a graph rewrite language based on PROGRES' | 18:56 |
| kanzure | Still not sure how to do this. Let's say you have 10,000 "CAD" files each doing various permutations and such. Where does the ANN come into play? You could spend a few thousand iterations tweaking parameters on a single file if you had to .. | 18:57 |
| kanzure | 'An alternative idea would be to let the ANN do the prediction of the success of a design running through the simulator. An application of a rewrite rule to the ANN would then cause it to make a possibly better or worse prediction of the results of the simulation. ' | 19:26 |
| kanzure | (well, that's really my text) | 19:26 |
| kanzure | would that be a useful approach? | 19:26 |
| ybit | " [17:16] <bkero> ANN? " http://en.wikipedia.org/wiki/Artificial_neural_network | 20:21 |
| kanzure | ybit: Did I miss that message of bkero's? | 20:21 |
| kanzure | and what was it in response to ? | 20:22 |
| kanzure | Oh, you're replying to it. | 20:22 |
| ybit | you didn't have a response :P | 20:22 |
| ybit | oh, for your question, not sure if you want my response | 20:22 |
| kanzure | hm? | 20:22 |
| ybit | " [18:26] <kanzure> would that be a useful approach? " | 20:22 |
| kanzure | Heh. | 20:23 |
| kanzure | Do you know? | 20:23 |
| ybit | i do not | 20:23 |
| ybit | unfortunately | 20:23 |
| kanzure | I think I'll just let the ANN predict the results of the simulation of the given model. The input to the ANN has to be "the given model" but you're not just going to dump YAML or CAD into it. You want to dump whatever makes a particular design unique out of the tens of thousands of other generated designs, as the input to the ANN. Now, each time that you generate a new 10k set of designs for some problem, you have to start working with fresh ANNs. | 20:25 |
| ybit | btw, did you ever happen to finish downloading http://www.med.wayne.edu neuropsychopharmacology vids from superkuh? | 20:27 |
| kanzure | No, the download for both fedt and I broke at the same time. | 20:28 |
| kanzure | What makes one design unique? Usually the connectivity between the components, possibly also the parameters for instantiating each part (i.e., if a parameter to a 'nail' is the length of the nail ..). But how do you express connectivity information to inputs of ANN? That doesn't sound right to me. | 20:29 |
| kanzure | Huh, what if it's the permutation ID of the particular design? | 20:33 |
| kanzure | Since it's generating them iteratively .. | 20:33 |
| kanzure | perhaps that's not a good idea. | 20:33 |
| kanzure | Maybe I'll let this be variable -- the input will be the ID of the generated graph; sometimes it will be the seed of the "random number generator" to generate the graph, etc. etc. Then it would be easy to test which ones will tend to have ANNs that accurately predict the simulation results. | 20:34 |
| kanzure | Simultaneously tuning both the type of input to the ANNs as well as the generation of ANNs for problem sets (via picking good 'rule sets' and 'graph grammars' to apply (rewrite rules)) is computationally ridiculous. But only I'd have to tune and test the inputs to the ANNs, and then everyone else would be testing the ANNs for (their new) problem domains. | 20:37 |
| kanzure | unless nsh might have a few ideas | 20:37 |
| kanzure | cough? | 20:37 |
| -!- Nade [n=lixasd@cpc2-clif5-0-0-cust516.nott.cable.ntl.com] has quit [] | 20:42 | |
| kanzure | When in doubt, make it a variable. Hurrah. | 21:13 |
| -!- willPow3r [n=will@cpe-66-75-6-181.san.res.rr.com] has quit ["Leaving"] | 22:08 | |
| bkero | $i = 4 | 22:23 |
| kanzure | Hm? | 22:23 |
| kanzure | the calculus professor blurted out during his lecture on series (yawn - been here, done that) that he had once attended a lecture of the great Conway, apparently he was demonstrating his tubes. | 22:28 |
Generated by irclog2html.py 2.15.0.dev0 by Marius Gedminas - find it at mg.pov.lt!