Advanced Cell Classifier

[Horvath Peter]

Dear All,

I am newbie on the list and happy to join. I am working in Zurich ETH
RNAi Image-based Screening Center as a computer scientist.

I would like to propose you the open-source program I am working on. The
program is called Advanced Cell Classifier (ACC) and designed to easier
analyze HCS data using machine learning techniques. For more information
you can see: http://acc.ethz.ch or you can write me an email.

Have a nice day,
Peter

Peter Horvath, Ph.D.
Light Microscopy Centre
ETH Zurich
m: peter....@lmc.biol.ethz.ch
w: http://www.lmc.ethz.ch/People/PeterHorvath
w: http://acc.ethz.ch
t: +41 44 633 34 94 / +36 70 7087590

[Jonny Sexton]

Hi Peter,

Looks interesting.

Are you using a support vector machine for the classification?

Best,

Jonny

********************************************************
Jonathan Z. Sexton, Ph.D.
Assistant Research Professor
Biomanufacturing Research Institute and Technology Enterprise (BRITE)
North Carolina Central University
http://brite.nccu.edu, jse...@nccu.edu, 919-530-6251
********************************************************

On Aug 3, 3:07 pm, "Horvath Peter" <peter.horv...@lmc.biol.ethz.ch>
wrote:

> Dear All,
>
> I am newbie on the list and happy to join. I am working in Zurich ETH
> RNAi Image-based Screening Center as a computer scientist.
>
> I would like to propose you the open-source program I am working on. The
> program is called Advanced Cell Classifier (ACC) and designed to easier
> analyze HCS data using machine learning techniques. For more information

> you can see:  http://acc.ethz.chor you can write me an email.

>
> Have a nice day,
> Peter
>
> Peter Horvath, Ph.D.
> Light Microscopy Centre
> ETH Zurich

> m: peter.horv...@lmc.biol.ethz.ch

[Horvath Peter]

Dear Jonny,

Thanks. As default I use the built in Matlab neural network since this performs well and fast, but from the beginning I built in the libSVM package which works also fine but not as good as NN. Since a while I also integrated WEKA, which offers around 40-50 different classifier and surprisingly (or maybe not :) ), there are few which gives really excellent result (I would emphasize the random tree classifier family and few boosting algorithms). Right now I am working on a function which helps to automatically choose the best method for the given data set.

Cheers,

Peter

Peter Horvath, Ph.D.
Light Microscopy Centre
ETH Zurich

m: peter....@lmc.biol.ethz.ch

w: http://www.lmc.ethz.ch/People/PeterHorvath
w: http://acc.ethz.ch
t: +41 44 633 34 94 / +36 70 7087590

[Jonny Sexton]

Hi Peter,

I have a very subtle phenotype that I am trying to characterize, and
have a good training set from a dose-response of a particular
compound. I have been using a partial least squares regression
against the dose-response to characterize the change in phenotype and
then transform the screening data, but I would be interested to see if
the NN can be trained to identify hits automatically.

What version of Matlab does your software use? How can we try it out?

Best,

Jonny

********************************************************
Jonathan Z. Sexton, Ph.D.
Assistant Research Professor
Biomanufacturing Research Institute and Technology Enterprise (BRITE)
North Carolina Central University
http://brite.nccu.edu, jse...@nccu.edu, 919-530-6251
********************************************************

On Aug 4, 3:04 pm, "Horvath Peter" <peter.horv...@lmc.biol.ethz.ch>
wrote:

> Dear Jonny,
>
> Thanks. As default I use the built in Matlab neural network since this performs well and fast, but from the beginning I built in the libSVM package which works also fine but not as good as NN. Since a while I also integrated WEKA, which offers around 40-50 different classifier and surprisingly (or maybe not :) ), there are few which gives really excellent result (I would emphasize the random tree classifier family and few boosting algorithms). Right now I am working on a function which helps to automatically choose the best method for the given data set.
>
> Cheers,
> Peter
>
> Peter Horvath, Ph.D.
> Light Microscopy Centre
> ETH Zurich

> m: peter.horv...@lmc.biol.ethz.ch
> w:http://www.lmc.ethz.ch/People/PeterHorvath
> w:http://acc.ethz.ch
> t: +41 44 633 34 94 / +36 70 7087590
>
>
>

> -----Original Message-----
> From: highc...@googlegroups.com [mailto:highc...@googlegroups.com] On Behalf Of Jonny Sexton
> Sent: Tuesday, August 04, 2009 8:58 PM
> To: High Content Imaging
> Subject: [High Content Imaging] Re: Advanced Cell Classifier
>
> Hi Peter,
>
> Looks interesting.
>
> Are you using a support vector machine for the classification?
>
> Best,
>
> Jonny
>
> ********************************************************
> Jonathan Z. Sexton, Ph.D.
> Assistant Research Professor
> Biomanufacturing Research Institute and Technology Enterprise (BRITE)

> North Carolina Central Universityhttp://brite.nccu.edu, jsex...@nccu.edu, 919-530-6251

> ********************************************************
>
> On Aug 3, 3:07 pm, "Horvath  Peter" <peter.horv...@lmc.biol.ethz.ch>
> wrote:
> > Dear All,
>
> > I am newbie on the list and happy to join. I am working in Zurich ETH
> > RNAi Image-based Screening Center as a computer scientist.
>
> > I would like to propose you the open-source program I am working on. The
> > program is called Advanced Cell Classifier (ACC) and designed to easier
> > analyze HCS data using machine learning techniques. For more information

> > you can see:  http://acc.ethz.choryou can write me an email.