Name_of_posting_person: Anselm Vossen
Type_of_posting: I am commenting on a previous posting
Date: November 28, 2005
Time: 11:11:55 PM
MessageJust a comment about Lilianas Track structure:
Maybe we should make a definite plan together, there are many important things
that I would like to see included in something like a "Introduction to ..." session like Feature selection,
simple Classification schemes like Bayes, things pertaining to different classifiers like overfitting.
So maybe we could have a little bit of coordination...
Cheers,
Anselm
Name_of_posting_person: Liliana Teodorescu
Type_of_posting: I am commenting on a previous posting
Date: November 14, 2005
Time: 03:57:40 PM
Message Dear friends,
Because Jarek Przybyszewski is interested to join the track on Artificial Intelligence in HEP, as he
let me know in an email, we modify a little bit out first proposal. The new structure of this track,
as we see it now, is:
Lesson 1: Introduction to Artificial Intelligence (Liliana)
Lesson 2: Machine Learning (Jarek)
Lesson 3: Neural Networks (Liliana)
Lesson 4: Support Vector Machine (Anselm)
Lesson 5: Evolutionary Algorithms (Liliana)
Still time to include your ideas as well!
Best regards,
Liliana
Name_of_posting_person: Marek Szuba
Type_of_posting: I have a specific topic to propose
Date: November 02, 2005
Time: 11:19:57 PM
MessageMyself I'd like to propose an overview of challenges related to constructing such large-scale and complex devices as high-energy physics detectors, with some possible solutions. As a member of the ALICE Detector Construction Database (DCDB) team I have been working in the field for quite a few years; I believe our system is the first one of this kind which attempts to cover all aspects and stages of the process, from construction of the most simple components through testing taking place at various stages of assembly to the assembly and cabling of the whole detector, however there have of course been partial attempts before and we (as well as I myself) are familiar with some of them. The system involves several variants of the interface: manual one for user, automated one for fast insertion of test data directly from test-bench software, a mechanism for handling bar-coding standard used for tracking all the components... Last but not least, the structure of the database itself as developed by our primary DB specialist is quite unique, featuring a high level of genericness which allows for using a single system for dramatically different needs of different subdetectors.
Furthermore, it would be great if I could somehow tie my involvement in parallel computing into the topic proposed by the other Marek - perhaps more from a physicist's point of view? I suppose a part of this would involve expanding on the topics skimmed over by my presentation at the CSC.
Name_of_posting_person: Vijayalakshmi Sundararajan
Type_of_posting: I have a specific topic to propose
Date: November 01, 2005
Time: 09:43:12 AM
MessageRegarding the iCSC2006, I think about a presentation on various kind of Software Testing and Commercial Testing tools (I hold a British Computer Society certification for Software Testing, and i have worked in some industry testing projects as well).
I can present about
- Functional Test
- Regression Test
- Integration Test
- Performance Test
- Stress Test
- Load Test
- Black Box and White Box Testing Techniques (Branch Testing,Equivalence Partitioning,Linear Code Jump)
- Static and Dynamic Testing
- V-Model of Testing
- Psychology of Tester and Developer
- Types of Testing tools (Requirement Testing tool, Test Design Tools,Test Data preparation tool,GUI,character based,Load,Comparison)
- Why and How Testing is very important for software development.
- Test Management
- Mercury Win Runner, Load Runner
I presently work on "Nightly Builds System and Testing" for CMS experiment. I can add this as well.
Vijayalakshmi Sundararajan
CERN, PH-UCM Group
Name_of_posting_person: Liliana Teodorescu
Type_of_posting: I am commenting on a previous posting
Date: October 31, 2005
Time: 10:20:47 PM
MessageHello Jaroslaw,
You might be pleased to find out that machine learning IS “directly connect with the area of interests for the CERN people” (I’ve used your words here!) Physicists are using machine learning techniques in data analysis and are, for sure, interested to learn more about them. In fact, many of the algorithms that I proposed in my previous email are also used in a machine learning style. I think your proposal for a machine learning lecture can fit well in the track that I proposed
(named there Artificial Intelligence in High Energy Physics) and I would like to invite you to join Anselm and me in developing this track.
Best regards, Liliana
Name_of_posting_person: Jaroslaw Przybyszewski
Type_of_posting: I have a specific topic to propose
Date: October 31, 2005
Time: 08:19:17 PM
MessageHi,
I would like to propose two topics from two different areas of computer science for the iCSC: 1) I have an idea of making lecture on database design, but I'd like to focus on different structures in Oracle that can store the data: tables, index organized tables, partitions etc. So it would be rather for users that have some experience in databases and database design. I could also do a "preparation" lecture - database design principles, introduction to databases, foundations of SQL. 2) The second topic I want to propose is rather not connected with CERN research directly. I would like to present foundations of "machine learning". The lecture would contain short introduction about machine learning, I would explain what are the reasons for "teaching" machines, when it can be useful and maybe if there will be enough time I would present few basic algorithms. In general the first topic would be different from the lecture prepared by Zori, because I don't want to do a full course on databases, but rather focus on some Oracle specific aspects that can improve database efficiency and we often forget about them. The second lecture is not directly connect with the area of interests for the CERN people, but this is a very new attitude in computer science and I think it could be very interesting.
Best regards for All of You!
Name_of_posting_person: Bobby Gomez-Reino
Type_of_posting: I have a specific topic to propose
Date: October 24, 2005
Time: 06:13:49 PM
MessageHi all,
Just a brief idea, I would like to propose slow control as another topic.
Cheers,
Bobby
Name_of_posting_person: Liliana Teodorescu
Type_of_posting: I have a specific topic to propose
Date: October 19, 2005
Time: 07:33:34 PM
Message Dear friends,
During the two weeks in Saint Malo we had very nice discussions in which many of you expressed interest in various computing algorithms and their applications in Particle Physics. Some of you are even working (or worked in the past) with such algorithms. This determined me to propose here a track with this topic.
The main idea of the track would be to discuss algorithms less commonly used in Particle Physics and to show how they are applied to typical problems from this field. I see that as an inspiration and encouragement for particle physicists to explore new algorithms and as an invitation for computer scientists to propose other powerful algorithms developed in their field.
Taking into account the spectrum of algorithms for which I noticed interest, I propose "Artificial Intelligence in High Energy Physics" as the name of the track and the following structure:
- an introduction session (1 hour) that will cover: definition of Artificial Intelligence (AI), short description of general classes of AI algorithms, typical problems where they can be applied.
- few sessions, each one dedicated to a certain type of algorithms that will cover:
a more detailed description of the algorithms, their performances in solving different type of problems, examples of their applications in Particle Physics.
As type of algorithms I would suggest:
- Support Vector Machine (SVM) (1 hour)
- Evolutionary Algorithms (EA) (1 hour)
- Neural Networks (NN) (1 hour)
-others (wavelet, fuzzy systems or any other idea you might have) (1 hour).
Anselm Vossen is particularly interested in SVM, me in EA and the introductory lesson.
We can cover the other topics as well but it would be nice if more of you will contribute.
If you are interested in this track just let us know and we can adjust it to include your ideas
as well.
Best regards, Liliana
Name_of_posting_person: Marek Biskup
Type_of_posting: I have a specific topic to propose
Date: October 17, 2005
Time: 01:51:19 PM
MessageFor the last one-two years we can see that the Moore's law concerning the CPU frequency doesn't hold any more. The biggest CPU manufacturers are trying to keep up with the Moore's law by creating more sophisticated architecture and using multi-core CPUs (several CPU units on the same chip). Intel and Amd have already
announced releasing a double core CPU for home usage in early 2006 and quad-core ones in 2007. Intel predicts that in the years 2006-2012 we will have a multi-core CPU era with even 32 CPU's in a simple desktop or laptop and many-core era from 2010 on with massively parallel applications running on hundreds of CPU cores, some of the cores specialized for
certain tasks such as speech recognition.
To be able to gain from the technology progress we will have to design parallel applications running on many CPU cores at the same time. I would like to propose a lecture (or a track if somebody else would like to join in - there is far too much to say for a single lecture) named "Designing
Parallel Applications" which will introduce students to this kind of
programming. The main topics I'd like to present are:
1. Introduction to parallel processing (motivation, current hardware).
2. Message passing (send, receive - blocking and non-blocking).
3. Basics of concurrent programming (mutexes, monitors, conditions, semaphores).
4. Multithreading and RMI in Java.
5. Multithreading in C++.
6. PROOF - parallel data analysis with ROOT (this actually might be a separate
lecture).
Name_of_posting_person: Marek Biskup
Type_of_posting: I have a specific topic to propose
Date: October 17, 2005
Time: 12:40:58 PM
MessageThe most popular software for end-user data analysis in HEP community at the moment is the ROOT framework. It contains tools for rapid histogramming and fitting as well as far more advanced methods which include c++ scripting, object IO and
parallel data analysis. I would like to give an introduction to ROOT and present its features which are
interesting for physics (histograms, fiting, math library, visualization) and computer
scientists (CINT - the C++ interpreter, reflection, object IO). I'd like to spend some time on PROOF (Paralled ROOT Facility) which is
becoming more and more important as the data volume is getting higher and higher. I consider giving a short (5 minutes) demo of ROOT and maybe PROOF.
Name_of_posting_person: Artem Harutyunyan
Type_of_posting: I have a specific topic to propose
Date: September 29, 2005
Time: 01:01:19 AM
MessageHi there,
I would like to propose a topic concerning security issues in Grid. We (me and Arsen Hayrapetyan) have created and are currently maintaining Certification Authority (CA, http://www.escience.am/ca of Armenian e-Science Foundation (ArmeSFo, http://www.escience.am which enters European Grid Policy Management Authority Group (EU Grid PMA) since December 2003. It would be interesting to talk about X509 Certificates (what information is kept there, what are extensions, how validity of certificate is verified etc.), proxy certificates (why do we need them, how do we create them etc) as well as to talk about algorithms for digital signing and encryption (symmetric and asymmetric encryption, usage of big prime numbers etc). It would also be nice to share experience of running a CA and organize hands on exercises with some demos (how to generate certificate request, how to sign it , how to verify it, how to extract info from a certificate, how modification of a certificate is done, how to extract prime numbers which are 'inside' the certificate, how to digitally sign files and emails, how to encrypt data and emails etc).
Artem.
Name_of_posting_person: Yushu Yao
Type_of_posting: I have a specific topic to propose
Date: September 28, 2005
Time: 07:28:18 PM
MessageSeveral Topics I think of, they are just ideas, not the formal proposal.
Software Related:
1. Creating Cross platform Software.
It becomes more and more important to make your application run on more platforms with minimum modification. And it’s even more useful to run your application through GRID. By keeping in mind the idea of transplantable application when developing, and by doing the development in the correct way, one can produce better transplantable applications with less pain.
Physics Relate:
2. Introduction to Event Generators.
This is a more physics related topic. Now there are all kinds of event generators/process simulators, they are focused on the different processes in different conditions and could be used for different applications. Pythia, Tphic, Lpair, Tomgam, Pandora, ... It will be good to make them organized and introduce them to the physics users. Which processes, what conditions, where to find, how to use, etc…
3. Starting GUIDE of Root.
From the basics, introduce Root Analysis Software to physics users. I don’t know if it has been included in CSC before, if not, as it's one of the most popular analysis software, it’s good to introduce it to the newcomers and the Fortran users.
Name_of_posting_person:
Ozgur.Cobanoglu@cern.chr
Type_of_posting: I have a specific topic to propose
Time:05:54:46 PM
Message
Hello dear friends,
I would like to propose "Data Quality Monitoring (DQM)",
seemingly as the starting item in the list :) This is what I do for ALICE
withing the DAQ group. It is a very practical study with lots of colorful
pictures in 2 and 3D. The detector people would like to see, on their
screens, what their detector "sees" on-line or off-line during the beam or
cosmic or whatever tests.
For that purpose, the DQM tool (in our case it is
MOOD for Monitor Of On-line Data & Detector Debugger) takes the data on-line
or off-line, decodes it and produces the plots that are required for the
user to decide whether the detector is in its good "MOOD" or not. DQM tool
"re-constructs" the event in a very fast manner, without running any
sophisticated/timeConsuming routins like cluster finding algorithms.
If a group of people who would like to create a
track (series of lectures) to which my proposal could contribute, I would
like to be a member of it. Since this is one of the really very practical
needs in our field, I think that it would be good to give people an overview
on how we (do not :-) ) handle this type of problems.
Cheers,
Özgür
Name_of_posting_person: François Fluckiger
Type_of_posting: Initial welcome message
Date: September 28, 2005
Time: 03:52:48 PM
MessageWelcome to all CSC2005 participants. I hope you enjoyed our school in Saint Malo and you will be willing to suggest topics through this forum.
François, School Director.
|
