Streaks of thought: Streak 12

Being aware that we can't understand everything is a sign of true scientist. Not doing anything because of that awareness is a sign of true idiot.

Exploiting structures

In one of my previous posts I mentioned that "abstraction" or "pure" is relative and is a continuous spectrum based on ones imagination capabilities. People like balance between surprises and predictability. This is because it gives non-triviality more value that lets our most basic ego survive in a modern way. The more the balance is off the more frustrating one becomes. Too many surprises or too much predictability can be frustrating in different ways. Working on "real-world" problems can be frustrating because of too messy structure and less predictability. Theorists tend to be easily frustrated by this and choose to work where things are more predictable. They work by putting a "nicer structure" (e.g. black-box model for query complexity) onto the real-world and exploit that structure in a non-trivial way. Typically since these are essentially imaginary structures one needs higher mental abilities to be able to produce. Engineers on the other hand tend to work on real-world structures directly as they can withstand more frustrations with less predictability (usually because of monetary rewards). Of course monetary rewards are proportional to the amount of work on the real-world structures. An important point to note is that both use "principled tools" for exploiting the structures. It's just that engineering tends to be more lazy (demand driven) than theory. Both are "supposed to be" scientific enterprises and essentially all human beings like exploiting structures which leads to my following paragraph.

Through out my blog you can notice hidden and explicit inclination towards capitalistic philosophy with materialism. Capitalism is based on simple premises and sophisticated creativity while communism is based on sophisticated assumptions and minimal creativity. Here I mean creativity that can be objectively verified. According to me objective verifiability is one of the major steps in human progress. This is what that led to modern standards which clearly can support unprecedented human population. A simple example is that in ancient days "killing and conquering" was an honor which can not be justified except for lack of objective materialistic understanding. Materialistic modern way can be very useful in reducing domination by race, killing etc. Since people like exploiting structures people tend to build structures of their own to work on them. Materialism can be a very useful guide to focus on real structures (including the "nicer structures" I mentioned in my last paragraph) instead of intangible unverifiable imaginary structures. Since verifiabilty is inherent in capitalism and materialism things can not go unchecked while this is not the case with communism.

Streaks of thought: Streak 11

Some jobs are boring because they are not creative and are repetitive. Well, every job is repetitive. Boredom of a job is proportional to the degrees of freedom given to take responsibilities. If we have more number of degrees and more freedom the space of possible ways of handling responsibilities just becomes larger so that we can "forget" the repetitions because of limited memory or knowledge. That's why having proportionally limited knowledge can reduce boredom!

Means and ends

About a year ago I had a post on balance motivated by self-characterization using souls and bodies. As one grows older one masters that part but new challenges come up like dealing with balance among other souls and bodies since our progress is defined inherently based on efficient interdependence. Typically the differences among souls part is very less and very trivial to handle except for the fact of bodies interfering which actually makes the whole life non-trivial and interesting (assuming there's no trivial impossibility either).

One essential difference between bodies is the gender. Like struggle between soul and body demands for balance, struggle between the gender characteristics also calls for balance. As I discussed in my earlier post we need awareness of the characteristics to realize the balance in a pure capitalistic way. The reason I like capitalism is it builds on simple premises but complicated (non-trivial) and creative proving abilities while communism builds on complex premises and simple proving abilities.

The main characteristic difference between females and males is that estrone drives the means while testosterone drives the ends. Now which is important, means or ends? Every kid probably grows up listening to the anecdote of "the wood-cutter sharpening his saw for better productivity". Without means we cannot get to ends and without ends there's no point in having means. So both have equal importance in a cycle of progress.

Once Richard Beigel told me "A good theorist does not focus on too many problems at one time." Divide and conquer is a very good strategy. Evolution also took this strategy in dividing the fuel that can focus on two equally important issues for progress and hence we have the two genders and in any divide and conquer a very important part is to put the conquered parts together and that's why both females and males have estrone and testosterone to allow for some level of empathy that can lead to actual full results.

Learning from United States of America

When I was in my undergrad in India sometime in 2001-2002, our college Principal Narasimha Reddy said "Go to America and learn the work culture". Recently in 2007 at a workshop on "Computer Vision applications for developing countries" that was in conjunction with ICCV in Rio de Janeiro Brazil, Ramesh Jain gave a talk in which he mentioned how the research in developing countries like India is aimed at working for USA since most researchers focus on publishing in conferences, journals run by USA which address problems pertinent to socio-economic situations of typical developed countries.

OK. The longer I stay in America the deeper my appreciation of America grows. What can we (developing countries) do to really learn from it? Ideally we want to learn the underlying process that generates the American success. We typically use observed data to learn the process: an inverse problem. We need data that can truly characterize American success. If we use wrong data no matter how strong our learning algorithms are, we are bound to learn wrong concepts. Typical problems that we run into in learning are that the data acquired is too noisy and even corrupted by unwanted signficant processes. So to be efficient in learning from USA we need data as little corrupted as possible. To acquire such data:

• One needs to have a first hand experience of life in USA.


• One needs to be open and courageous so that he actually collects the data.


• Should interact with atleast a few elite Americans not necessarily personally but through different media like reading articles, blogs and working etc.


• I say elite because they minimally corrupt the data. And if you encouter a few of them it's good enough to learn the ground truths and you can peel off layers when interacting with average or below average Americans.


• Check objectively if you are able to learn. A good test is to check if you like America more than just for money and comforts. Understand that I am not preaching money is not important I am just saying it is not the causal variable for deep dynamics. It's how we use money that matters.


• Periodically "smooth" your data by re-interpreting past experiences.

Just a side-note. Nobody is perfect. America has its faults but it is actually a worthy leader in doing what matters, that is struggling for ensuring longterm human survival. Doing nothing is trivial and doing things by being all powerful is also trivial. What is non-trivial is understanding the dynamics with constraints.

My expressions in last paragraph are influenced by my regular reading of Scott's blog. Especially the "trivial/non-trivial part" and "the doing what matters thing"s.

Note: I edited this post so many more times than any other posts after publishing.

Survival using science

Pure math is not very useful. Pure mathematicians publish results among themselves and never really get used. Applied mathematicians and engineers make the world.

We hear such statements from time to time in academia. Typically one perceives something to be useful if one can ensure survival through it. It's true that not many can survive in pure mathematics because it is so damn tough. Here I should mention that typically abstractions higher than one's level of imagination are considered "pure" by him. Pure is rather a spectrum of higher abstractions. In fact the higher the abstraction the higher are the chances that the results can survive through odds or in other words stand the test of time. I encourage to read this article and this one to get a nice perspective on math.

Conferences are a good way to measure the progress and importance of a field. For e.g. conferences on aritificial intelligence kind of died but several of them are budding again. Hence participating in conferences is a big treasure to learn the dynamics of a field. If you can actively participate in conferences of a particular field it means you possibly can be fit in the survival sense in that field. You can always learn not only from active participation but also from passive participation. By passive participation I mean being mostly a spectator. Conferences' radiance can be measured in different dimensions like by rigor (typical pure math conferences), immediate impact (engineering conferences), deeper impact (theoretical conferences), commercial impact (industrial conferences). Even though you survive using one class of conferences it always helps to gain edge benefits by being aware of the dynamics of other classes.

After all people in academia struggle to survive primarily using science. It's important for one to have broader awareness if he wants to survive the real significant dynamics. The deeper or steeper one's understanding is the higher are the chances of his survival in times of crisis. Rationality is sufficient when things are smooth, rationality with rigor is necessary when sailing through rough waters.

Streaks of thought: Streak 10

Those religions which are designed primarily to aid people live happily are not very useful in realizing GOD and hence they seem silly some times. Religions which are designed primarily to realize GOD and use happiness and success as tools, are bound to lead. Such religions only demand to be successful so that there is enough fuel (life) to pursue realization of GOD. This way we are at a much higher abstraction and lot of sayings like "don't take success to your head", "don't hurt others" etc. are automatically explainable.

My only favorite is Judaism! For e.g. it talks about sex much differently than other religions since sex should not distract us away from our path to GOD. If a religion is not designed to keep people on track happily it's going to be face trouble.

Thank GOD!

Why do we thank GOD for good things and cannot blame HIM for the suffering and bad things happening?

This is a common question posed by atheists. GOD is all-powerful but makes the "game of life" interesting by playing the role of honest prover. If good things happen (convinced of a new theorem) we thank HIM for being an honest player. If bad things happen (unconvinced of a theorem) then we cannot blame HIM as the fault lies on the side of the verification though several times not obvious. Blaming is a sign of tiredness of playing the game and lack of sportive spirit. And do you like players who lose temper in games?

A related post.

Responsibility and credit

Events in life result because of deterministic and non-deterministic sequence of actions. Taking credit for deterministic actions is an easy thing and hence not really interesting. Here I am talking about personal credit, not credit given by society because of being the medium for events. According to karma we can take credit for non-deterministic actions but it is usually hard for most to be good accountants. It takes experience, courage and many more hard qualities which grow by practice to be good accountant.

A subtle but important point to realize is that we are only playing a role of verifier in the system (Nature if you may think so) where some sequence of actions result in events. Since we do not explicitly know who plays the role of prover we cannot take "complete" credit (personal) for events even if we believe in karma. Believing in G-d (I am not talking about belief in after-life or being religious etc., which are only some efficient tools) is probably the optimal choice for the role of prover, who plays a key role in the events.

One important point is that we need to assume the prover and verifier are both honest for the completeness requirement of the system. It is necessary that the verifier is honest for the system to be sound. One can imagine that if there were an all-powerful bad-guy then honest verifier will not allow him win!

Principle of abundance

Modern science laid the universal foundation for rational materialism. Though materialism might be incompatible with several religious beliefs true materialistic rationality can not mislead us in our search for understanding the universe. It is important that it should be true and rational materialism. In fact given the complex structure of the society our survival is currently based on, it is probably the optimal choice to take on. This is because we can then work on complexity of the universe since we can concretely measure (and hence share) our understanding of the universe in terms of "material". To efficiently study the complexity the asymptotic view of amount of resources needed and several "on average" analyses like amortized analysis have shown to be very useful for e.g. in understanding nature of computation which has contributed to understanding universe.

The key to be effective in studying complexity is to be smart enough to care about details and mature enough to have a macroscopic perspective. If a complexity theorist gets too involved in non-asymptotic details he might not be able to make any real contribution. If he is not careful enough with the details of his proofs and assumptions he will be wasteful of resources. Working on non-asymptotic details like creative engineering is an evolving profession like that of practicing law, medicine, programming where as more and more complexity is revealed it becomes more and more important. To work based on asymptotic behaviors it's important to have an abundance perception of the universe in terms of time, life, opportunities, possibilities etc. along with finiteness. If we do not work in finitistic set of assumptions things might turn out to be too abstract to be useful some times.

It is important to understand that the abundance is finitely measurable and is not free which makes the process interesting and safe.

Streaks of thought: Streak 9

Proving lower bounds for complexity (defined in terms of resources needed) of problems is known to be hard. Lower bound on a resource is the minimum amount needed to solve the worst of the instances of a problem. Hence usually worst case analysis is needed to prove lower bounds. The intention of such worst case analyses is mainly to install signs to control traffic of exploration for saving energy and being effective.

Such effort should not be confused with immature and dumb "but what ifs". Immaturity is usually flagged by non-productiveness of thought process. Dumbness is hard to be flagged especially because it's not as penalizing as immaturity for surviving.

MEXperience

I had learnt somethings from my experience building mex files for C-MATLAB interaction. Below I list some helpful points to keep in mind in cases of frustration:

• "mexing" in Linux/Unix requires a new setup of compilers.


• Make sure you have compatible gcc with the MATLAB version installed.


• Sometimes commenting lines of code using '//' might produce errors. Using '/*....*/' is safe.


• mexing on 32-bit Linux produces .mexglx executables and those are not executable on 64-bit Linux machines which need .mexa64 executables.


• mexPrintf statements can make your program really slow.


• Having too many "mx.." statements in the C-code might make the code run slower possibly because you are just repeating what MATLAB might already be doing without your C-code. Remember MATLAB does use some of its built-in C-code for speeding up certain operations. As an example if you have to handle structures in your C-code use "mx.." statements to pass parameters but use the "struct" variables in the C-code not mxArrays.


• Initially try to balance of how much of your code you want in C-style and how much in MATLAB. After you get familiarized with "mex" code try to code most of it in C and then use MATLAB as an "interace" for experimenting with your code.


• If building huge systems which you will probably do only after getting familiarized with "mex" forget MATLAB and use development tools like Visual Studio etc. for building and debugging then just add the interface. Debuggin on smaller projects can be handled effectively using mexPrintf which later have to be minimal.

Spiritual materialism

Recently I attended a spiritual retreat organized by Brahma Kumaris World Spiritual Organization at Peace Village in upstate NY. The retreat I attended was titled "My Self, My Work and My Life". This one was special because the theme was to discuss how to optimize your gains from life. The speakers and audience bounced around the ideas of how to "align" universal values (like to be peacful, loved, happy), thoughts and actions to maximize subjective gains and hence optimize overall gains.

Such exercises can provide us with insights that can help us to be creatively materialistic. It is not completely impossible to be creatively materiaslistic without spirituality but it can certainly help and make things lot easier. An analogical post that actually motivated the theme of this post is this.

Update: Added a picture I took at Peace Village to support my ending sentence of the post. 3:56 PM

Pasticide

Attachment to past (any kind, even good) is the biggest reason for procrastination. Procrastination is the biggest killer of opportunities. Opportunities are the biggest gifts of a societal system. Society is the biggest asset the humans have ever consciously created.

So try to use pasticides like detachment by focussing on productivity to optimize your contributions.

Beware of the fact that pesticides help in increasing the chances of reap but overusing might ruin the crop. So use pasticides carefully and maturely as you can be inspired from this post itself:)

JUNE 27, 2007. Corrected some stupid sentence constructs.

Streaks of thought: Streak 8

Thank you for shopping at our place! Thanks for your help buddy. Thank you very much for dropping me off here. Two or more parties involve in an act if and only if all of them have stakes in the act. Then why thanking? Reason for such thanking is not for the act itself but for sticking only to the act and hence helping in performing the act pleasantly to optimize everyones resources (energy, time, money). By the way all the resources that we really worry about in the current world are manifested in the form of money. Basic resources like food, clothing, shelter are not a serious concern in the age of abundance.

Tales off of tension

I came accross on the web something like this: Humans by definition have "spirit" and "body" and there is constant "tension" between the two parts. I think these tensions are the main cause for all the history of the humans. This tension is probably the hidden variable manifesting itself in the form of complex tales of human life. A "good" amount of tension is needed to hold the structure (otherwise there would be no history). If there are surges we see breaks in the structure. In fact how well the tension is balanced influences how stable the system is. According to Robin Hanson as mentioned in Scott's post academia is the least of the broken systems.

I think the reason that it's least broken is that this business is most exclusively handled by powerful elite and they usually tend to inherit "being more moral" spirit along with "being elite" genes and hence the breaks caused by fraudulent tensions are not a serious problem at all for this enterprise. There are those little breaks in the system because there are surges in tension between worrying about deductions and struggle to understand the assumptions.

Sprechen C MATLAB

Sprechen Sie MATLAB? This was on an ad-poster by Mathworks that I saw posted on Longin Jan Latecki's office door. No literate can deny how much Computer Science has influenced thinking styles and the pace of research in human quest for knowledge (genomics, bioinformatics, software simulations etc.). MATLAB is one product that has such outstanding influence on many many applied scientific areas (I call applied because I am not sure if MATLAB has led to any purely theoretical insights. By theoretical insights I mean ability to see non-trivial proofs.)

I have used MATLAB for several of my projects. MATLAB is an interpreted language and is known to be slower for 'C-style' programs. But in many cases you need to write blocks of code in 'C-style'. Since MATLAB allows you to interface with C-programs it's worth to know how to write C-programs that you can call in MATLAB without using files as communication (using files again slows down things doesn't it?)

In learning any thing new just don't let "But what if" trouble bother you that much. What if I want to pass pointers, what if I want to pass cell arrays, what if I want to return more than one variable from C etc. Well, you can do all these things but don't worry about all those at once before even getting started, deal one thing at a time. I borrowed this But what if idea from Scott's post about such problem in theoretical perspective which is much more serious! Now let's C how to start building the interface.

The key problem in building any interface is passing around parameters without confusion. The basic anatomy of a typical C-MATLAB file can be seen here. I have simple and clear demonstrations for passing scalars, arrays and structures. The README shows how to use the demos. I recently fixed a memory leak in "kdnn" function of Kd-tree software package for MATLAB written in C by Guy Shecther. You can download the "kdnn" function from here after downloading the original package. kdnn finds k nearest neighbors for a given d dimensional point.

Health: New challenges

After you push your health conscience to be at subsconscious level, what's next? You know we always need some thing in conscious mind to keep us busy. Infact the central assumption for the entire progress of mankind is that we consciously keep focusing on something new. Well let's stay focused on what's new to focus on, in the journey towards better health. Here are two reeaal challenging tasks you might want to focus on.

• Rock hard abs absolutely rock and are hard:
  In my opinion both huge and tiny people are trying to do the same end thing (it's just their pathways are different). Both of them try to look proportional and look the best they can. Initially when they start working out the sizes of the dresses are good inidicators for their progress. But after they start looking fine in dresses some start being content. But the naked truth is that they should aim for looking good naked. What are the challenges you have to face? The main challenge is toning muscles and especially the hardest part is toning abdominal muscles and even harder is to handle the lower abs. The main thing ofcourse is to get rid of the fat and build core muscles so that the abs look firm and in shape. My point is not to give advice on how to get that done as you can find lot of information thanks to Google and those who put the information up (Sample link that my friend and health guru Reagan gave me). My main point is to wake up those who are content just because they think they look good enough now when dressed.
  
• Look good underskin:
  Now what about those who already look good naked. Here's a pill that will wake them up too:) If you have fabulous abs and defined chest, biceps etc. you might still not look your best if you have troubles inside. Make sure you look sharp by solving puzzles and keeping your brain warm. Take time to meditate so you can give a peaceful presence whereever you are. Do YOGA type exercises to keep your deep-internal glands work properly. What I mean for e.g. is, you can keep your heart pump blood well by cardio-workouts but what can you do for pancreas and thyroid glands work properly. YOGA has some workouts for such subtle benefits which are very notorious for not motivating people until they get hit.

These two are new challenges that you have to focus on for progress in your journey.
Link to previous health post.

Streaks of thought: Streak 7

Based on several incidents like reading this and brain storming with other researchers trying to piggy back on results from theoreticians, statisticians and mathematicians I had the following streak:

Do not get held back just because you don't understand a result intuitively. Definitely continue to look at it closely until your intuition levels evolve to the required standard but not at the cost of not-using the result. Math has been much more powerful reasoning tool than intuition since intuition evolves much slower than objective reasoning. Another famous saying that goes along with this streak is "One century's wisdom is common sense of the next."

Updated: April 26, 2007. 9:54 PM

Universal language

Yesterday I watched CONTACT, a movie about a scientist obsessed about Search for Extra-terrestrial Intelligence (SETI) and contacting some alien in the form of her dead dad on VEGA. The story also has a theologian who is not against Science but questions the privilege that Science gets. There is one line that is very nice. "...Universe is vastly huge...if we are the only ones in this universe..then it's awful waste of space..."

One thing that got my interest is the assumption that our Math is the universal language. Math is certainly standard language of Science on Earth. But why should we assume that it is universal. Even the purest of Mathematics on Earth depends on a set of assumptions like ZFC. What if ZFC is not valid in other parts of the universe? What if axioms of deduction are not valid? Perhaps this might be very important question that people already think about in that field. What might be important for CS on Earth is this. We often believe (for some reason) that aliens must be super intelligent than we are. And if so can they solve NP-Complete problems? Let's say that if they are able to solve then is communicating with them NP-hard? Is this the reason we are not able to communicate with them (if at all they are) in the first place. My conjecture is that the communication is NP-hard since the search space of mapping between their assumptions and our assumptions is close to be unstructured and exponential in size.

BTW it's a good long movie to watch.