Wednesday, September 29, 2004
Planning planning planning !!!!!
Its all about planning when it comes to research. I need to plan my research well. There are so many different ideas that I want to try and see run through as a prototype. Well, most of these concepts are related to one another. I need to lay out a plan to do these things one by one.
So I need to plan my research activity well. May be a timeline would be a good idea. Timeline need not be strict or detailed. It can just be a rough draft to outline main priorities or milestones.
I have the basic plan laid out that will involve publication in journal or in a conference in each semester of my graduation. Well, that said, is it something I control. No, I don't control the development speed of any idea or its implementation speed for that matter. In the middle of writing proposals, family and my actual primary research work, it leaves me no time to even develop new ideas I have. Leave alone implementing them.
It is as if I am a detective following several leads at the same time on the given case. Will I hit the right clues at the right time.. I don't know but I surely want to keep trying. Thinking never goes waste. I guess I should just start noting down those ideas and start following them as leads one afetr another in a pre-determined sequence. New things will pop up but I need to keep note of the existing ideas. Well so long......
So I need to plan my research activity well. May be a timeline would be a good idea. Timeline need not be strict or detailed. It can just be a rough draft to outline main priorities or milestones.
I have the basic plan laid out that will involve publication in journal or in a conference in each semester of my graduation. Well, that said, is it something I control. No, I don't control the development speed of any idea or its implementation speed for that matter. In the middle of writing proposals, family and my actual primary research work, it leaves me no time to even develop new ideas I have. Leave alone implementing them.
It is as if I am a detective following several leads at the same time on the given case. Will I hit the right clues at the right time.. I don't know but I surely want to keep trying. Thinking never goes waste. I guess I should just start noting down those ideas and start following them as leads one afetr another in a pre-determined sequence. New things will pop up but I need to keep note of the existing ideas. Well so long......
Sunday, September 12, 2004
Intuitive thinking
Last few days I have been thinking about ways to introduce new ideas in machine learning (AI) by the means of reasoning and knowledge representation. Deductive learning is very difficult but that is the way we humans think and process information. Now the idea is to make computers think the way we do but in more sophasticated or simpler manner. Roadblocks..?? well... there are many. First, we don't know much about our brain as we ought to in order to mimic it. Second, theories are limited by existing technologies and infrastructure to get it from paper into a prototype. It would be really cool to do information processing through means of reasoning , association and learning
Another thought I am working on past few days is of many different ways the research can hit a pause state in day to day life. For married guys especially, it could be a different challenge. My wife is a great support but when you have a house, and a family, you have some kind of responsibility. There has got to be a good balance of family and research in this process but the more you try to balance one, the other side is let loose. I believe it is that line one needs to understand and walk to be a good researcher and still have your family. I was told, Einstein was great scientist but was he a great guy to have a family..? I doubt it. To be honest, there is always this level of speculation and there is no perfect answer for it. To keep doing it and keeping on working on yourself is the only way out. Sometimes, I see these genius professors of my university. They have families and seem relaxed and comfortable instead of running around. I will be in those shoes some day..
Another thought I am working on past few days is of many different ways the research can hit a pause state in day to day life. For married guys especially, it could be a different challenge. My wife is a great support but when you have a house, and a family, you have some kind of responsibility. There has got to be a good balance of family and research in this process but the more you try to balance one, the other side is let loose. I believe it is that line one needs to understand and walk to be a good researcher and still have your family. I was told, Einstein was great scientist but was he a great guy to have a family..? I doubt it. To be honest, there is always this level of speculation and there is no perfect answer for it. To keep doing it and keeping on working on yourself is the only way out. Sometimes, I see these genius professors of my university. They have families and seem relaxed and comfortable instead of running around. I will be in those shoes some day..
Monday, September 06, 2004
Take a break
I am enjoying long weekend for labor day right now. Well, I have huge loads to research work lined up. I think I need some break from the work. I utilize long weekends (there aren't many) towards this.
Essentially even research is modular work just like professional projects. Especially prototyping part. I would think of research work into 4 phases. You can have these 4 phases for each problem, sub-problem and sub sub-problem in your research work. they are as follows:
Ok, coming to Min-Max algorithm, I talked to Nitin (My school friend) discussing this problem at length. I believe/propose that all documents are meant to be in individual cluster. Remember that clusters are suppose to have similar items clubbed together. But with different clusters comes the concepts of inter-cluster distances and intra-cluster distances. If we have only 1 cluster for all documents, notice that inter-cluster distance would be zero but intra-cluster distance would be really high that is indicator that even dis similar documents may have got clubbed together by mistake. On the other hand if we have 1 cluster per item to be classified, then we are looking at zero intra-cluster distance but now, the sum of distances of every cluster centre with other cluster centres (referred to as centroids) would be maximum. So we need to have balance of both. Idea is to have no/few parameters to drive clustering and automatically determine the ideal clustering looking at both the properties. But it is easier said than done. Now as I mentioned above, this is concetualization and now needs finer details to build a prototype and run experiments. Let us see...!!
Essentially even research is modular work just like professional projects. Especially prototyping part. I would think of research work into 4 phases. You can have these 4 phases for each problem, sub-problem and sub sub-problem in your research work. they are as follows:
- Initial conceptualization/theory of what you want to do.
- Some basic analysis/outline is followed by prototype development.
- Work with results of the experiments. Mark all different cases and leave no test case untested.
- Write a report on those experiences which may or may not be a research paper.
- I have to work on conceptualization in more details for the clustering problem. I am looking at clustering by random sampling or min-max algorithm. I will write about this in coming paragraphs.
- I have to work on the formula to calculate delta i.e. the one I have formulized after discussing with Dr. Cox of Mathematics department.
Ok, coming to Min-Max algorithm, I talked to Nitin (My school friend) discussing this problem at length. I believe/propose that all documents are meant to be in individual cluster. Remember that clusters are suppose to have similar items clubbed together. But with different clusters comes the concepts of inter-cluster distances and intra-cluster distances. If we have only 1 cluster for all documents, notice that inter-cluster distance would be zero but intra-cluster distance would be really high that is indicator that even dis similar documents may have got clubbed together by mistake. On the other hand if we have 1 cluster per item to be classified, then we are looking at zero intra-cluster distance but now, the sum of distances of every cluster centre with other cluster centres (referred to as centroids) would be maximum. So we need to have balance of both. Idea is to have no/few parameters to drive clustering and automatically determine the ideal clustering looking at both the properties. But it is easier said than done. Now as I mentioned above, this is concetualization and now needs finer details to build a prototype and run experiments. Let us see...!!
Thursday, September 02, 2004
Life is not easy
I have been thinking hard about the research sub problem at hand. I did come up with a way we cluster given data into meaningful groups. Each group will have data items that are similar to other data items in that group. What I realized today is that there already has been this work done and published in a reputed journal.
Now it means I better catch with my reading and not loose focus on what other research in my area is because I might stumble upon something that has already been done. This paper of the work I was thinking about was published in 1999 and that means I am 5 years late to think about that approach.
Most of the research work has to have a foresight where people can get to know about your research and they find it current. It is one of the critteria of finding it significant.
On other note, I have been a big fan of using CVS (Concurrent Versioning System) and know the importance of versioning and backing up your work. It also is a good measure of tracking changes and I have started extensive usage for my research work. My advisor is extremely happy using it in the entire research group and asked me set it up for the entire group. He also sent email to all of us asking to get CVSized in terms of our projects. Now for other research guys in my group who have never worked with CVS or who have never worked in professional environment or who have no interest in switching from their nice little nest they call Microsoft, it very difficult concept to grasp. It is unfortunate that I have been trying to promote and support good things happening in my research group.
I wish we all guys could go out for picnic, I wish we all group janta can discuss our research problem and help each other out. Mostly if you have friends from your country, you don't care for someone else who doesn't belong to your country. I respect all the countries and wish for everyone accommodated when there are discussions going on. I hate when people start talking in their local regional language in front of the person who has no clue what is being said. It is embarrassing for that one person to just keep looking at faces. I do take care about such an issues but people seem to be less sensitive towards others in this matter.
Anyways, this is suppose to be my research talk and not personal gossip.
BTW, I could get to the point at the end of the day today where I can distinguish my work and in fact may be I will come up with a better clustering approach than the already existing one. I am not giving just yet....
Now it means I better catch with my reading and not loose focus on what other research in my area is because I might stumble upon something that has already been done. This paper of the work I was thinking about was published in 1999 and that means I am 5 years late to think about that approach.
Most of the research work has to have a foresight where people can get to know about your research and they find it current. It is one of the critteria of finding it significant.
On other note, I have been a big fan of using CVS (Concurrent Versioning System) and know the importance of versioning and backing up your work. It also is a good measure of tracking changes and I have started extensive usage for my research work. My advisor is extremely happy using it in the entire research group and asked me set it up for the entire group. He also sent email to all of us asking to get CVSized in terms of our projects. Now for other research guys in my group who have never worked with CVS or who have never worked in professional environment or who have no interest in switching from their nice little nest they call Microsoft, it very difficult concept to grasp. It is unfortunate that I have been trying to promote and support good things happening in my research group.
I wish we all guys could go out for picnic, I wish we all group janta can discuss our research problem and help each other out. Mostly if you have friends from your country, you don't care for someone else who doesn't belong to your country. I respect all the countries and wish for everyone accommodated when there are discussions going on. I hate when people start talking in their local regional language in front of the person who has no clue what is being said. It is embarrassing for that one person to just keep looking at faces. I do take care about such an issues but people seem to be less sensitive towards others in this matter.
Anyways, this is suppose to be my research talk and not personal gossip.
BTW, I could get to the point at the end of the day today where I can distinguish my work and in fact may be I will come up with a better clustering approach than the already existing one. I am not giving just yet....
Wednesday, September 01, 2004
Requirements towards writing proposal
I am suppose to start working on proposal draft as I already have one conference publication by now.
I am tryign to gather information as to which format do we follow and also how to approach this. For any research there is one main problem (main area of research). That main area consists of many sub problems(each of them likely a publication). I talked to Anindo Roy yesterday. He is a PhD student here at UALR. According to him, we can take 2 routes for proposal defence.
1. One can form the advisory committe and put in the initial proposal immediately mentioning the significance of the problem and suggest what he/she plans to solve it. Offcourse there would be suggestions from the committee and then you go ahead with your research as discussed with everyone.
2. The second track which is rare or more difficult at times, is to talk to advisor about problem statement and its significance. Once that is established, start with survey, finalize few ideas and straight away get on with the research work. Publish few papers and then propose to the advisory committee about it.
In the later option, you risk being incomplete on wrong path of research if the advisory board feels that way. You also have to write exhaustive report in second case. You should be able to convince yourself and board with some preliminary results showing potential of the research.
On the other hand, first approach is rather easy and needs small writeup as proposal draft. But also the importance of the problem could be questioned. It is not worry anymore in second case as you already have a few publications to your credit in that field.
I am following second route and plan to publish one or two more papers before I actually submit proposal draft formally.
As for the problem of how to approach, I am still thinking about it. I need to be exhustive enough to explain each sub-problem in detail and also show some results in each case.
More later on this issue.
I am tryign to gather information as to which format do we follow and also how to approach this. For any research there is one main problem (main area of research). That main area consists of many sub problems(each of them likely a publication). I talked to Anindo Roy yesterday. He is a PhD student here at UALR. According to him, we can take 2 routes for proposal defence.
1. One can form the advisory committe and put in the initial proposal immediately mentioning the significance of the problem and suggest what he/she plans to solve it. Offcourse there would be suggestions from the committee and then you go ahead with your research as discussed with everyone.
2. The second track which is rare or more difficult at times, is to talk to advisor about problem statement and its significance. Once that is established, start with survey, finalize few ideas and straight away get on with the research work. Publish few papers and then propose to the advisory committee about it.
In the later option, you risk being incomplete on wrong path of research if the advisory board feels that way. You also have to write exhaustive report in second case. You should be able to convince yourself and board with some preliminary results showing potential of the research.
On the other hand, first approach is rather easy and needs small writeup as proposal draft. But also the importance of the problem could be questioned. It is not worry anymore in second case as you already have a few publications to your credit in that field.
I am following second route and plan to publish one or two more papers before I actually submit proposal draft formally.
As for the problem of how to approach, I am still thinking about it. I need to be exhustive enough to explain each sub-problem in detail and also show some results in each case.
More later on this issue.
