We’re going to make this into the place where scientists and engineers will go when they’re looking for reliable information on environmental problems, or ideas for projects to work on.

We’ve got our work cut out for us. If you click the link today — September 27th, 2010 — you won’t see much there. But I promise to keep making it better, with bulldog determination. And I hope you join me.

In addition to the wiki there’s a discussion forum:

where we can discuss work in progress on the Azimuth Project, engage in collaborative research, and decide on Azimuth policies.

Anybody can read the stuff on the Azimuth Forum. But if you want to join the conversation, you need to become a member. To learn how, read this and carefully follow the steps.

You’ll see a few sample articles on the Azimuth Project wiki, but they’re really just stubs. My plan now is to systematically go through some big issues — starting with some we’ve already discussed here — and blog about them. The resulting blog posts, and your responses to them, will then be fed into the wiki. My goal is to generate:

• readable, reliable summaries of environmental challenges we face,

• pointers to scientists and engineers working on these challenges,

and lists of

• ideas these people have had,

• questions that they have,

• questions that they should be thinking about, but aren’t.

Let the games begin! Let’s try to keep this planet a wonderful place!

Related

This entry was posted on Monday, September 27th, 2010 at 8:26 am and is filed under azimuth. You can follow any responses to this entry through the RSS 2.0 feed.
You can leave a response, or trackback from your own site.

I will probably post more on ‘how to use the Azimuth Project’ — and everyone is encouraged to ask questions here — but also, people should check out the How to page.

Great! (I’m in a hurry, have not yet worked with any wiki, etc., and will be mostly gone for this week… so: ) An Azimuth Project entry I (or you) will (should) be doing is “Carbon Negative Energy”. Got some links on my blog, could be more. The only source of carbon negative energy currently known is wood gas leaving behind char coal. John knows about it. But many engineers seem to disregard it. Being too primitive the idea seems to insult the tech ego. Even the “sustainable energy” business seems not to be interested. They prefer to sell solar and biogas, which, being not carbon negative is actually not sustainble – It’s a baby step in the right direction taken decades too late… That’s why I couldn’t wait till next week to mention this.

It was incredibly easy to do this. I just typed “Carbon negative energy” into the search box on top, and since a page by the name did not yet exist, I was given the option to create one with a single click of the mouse.

Then a place to type stuff appeared, and I typed some stuff in there. Nothing very interesting yet! Just the basic idea.

Then I clicked the “Submit” button at the bottom, and the page was created.

I hope you add more, Flori. To do so, just click on the link above, then click “Edit” at lower left. Then type in stuff. Don’t worry too much — it’s almost impossible to do anything that can’t easily be fixed.

The most important thing is to enter useful information. The second most important thing is to write clearly. Formatting and technical issues come last: the lab elves can clean up any mess you leave. So don’t use those issues as an excuse to put off saving the planet!

Of course, after you get into the habit of entering information in Azimuth, it’s good to learn some tricks for doing it well. For that, start here:

The article on Carbon negative energy is just a pathetic stub, but I just created two slightly more substantial articles on Florifulgurator’s favorite form of carbon negative energy:

I won’t be able to create every article that someone wants — you folks have to create them yourselves! But biochar and terra preta are incredibly important, potentially, and I happened to have some information about them at hand…

When everyone – tout le monde, as Tom Wolfe used to put it, meaning a relative handful of people, but everyone who supposedly matters – is saying something, it takes a real effort to step outside and say, wait a minute, how do we know that? It’s especially hard if you spend most of your time hanging out with other Very Serious People; I know that I myself have a hard time saying that people I know personally are talking nonsense, even when they are. The VSP effect is one reason smart bloggers, both on economics and on politics, have generally been a better guide to what’s really happening in America than famous reporters: their distance, their lack of up close and personal insights, is actually an advantage.

Then lower down:

This is what you need to know: important people have no special monopoly on wisdom; and in times like these, when the usual rules of economics don’t apply, they’re often deeply foolish, because the power of conventional wisdom prevents them from talking sense about a deeply unconventional situation.

Hmm. On the other hand we need genuine subject-matter expertise, which Krugman provides on Economics. On the other hand it is very easy for experts to be too narrowly focussed. I keep asking questions on his blog along the lines of “If world-wide demand were raised by world-wide government action to the level necessary to get unemployment down, then how much oil production would the economy need, and can we produce that much oil at the moment?”. My point is that (a) we need government stimulus directed to produce cheap energy to be an (imperfect) substitute for oil; and (b) we need to also soak up private demand (energy crisis bonds?) to prevent smashing (again) into the limit of oil production and crashing the economy. But am I right? As Krugman and others point out, things are very similar to Japan 15 years ago when there was no energy or other shortage. We are missing any way to think clearly about these things. Just different groups of people using lots of adjectives and doing lots of arm-waving (yes there are lots of numbers being thrown around, but their implications are far from clear).
Which brings us back to the question of who has a special monopoly on wisdom. I claim that the subject matter of mathematics is how to think clearly about problems. Yes mathematicians spend time thinking about unimportant problems that they just happen to be able to describe succinctly. And also trying to understand mathematics itself better. But the real problems that drive mathematics are in the real world. Inference is a universal aspect of clear thinking, and this has to involve Bayesian analysis and using maximum entropy to understand what we know before we look at the evidence and how the evidence modifies what we know [and I’m not saying these are easy tools to use]. But this doesn’t get us very far in understanding real world economic and environmental problems. I’d be rash to comment but I feel that the place to look has to be flows in configuration space, and the principle of Maximum Entropy Production will be the key to understanding that.

But the real problems that drive mathematics are in the real world.

Back when I was a pure mathematician I would complain about people using the world ‘real’ so often like this. Mathematics has its own reality with its own real problems which are beautiful and profound. Unfortunately us rats squirming in the muck of a suffering planet have our own urgent problems, and I’ve decided I need to think about those. These problems aren’t more ‘real’, but we’re stuck with them.

Inference is a universal aspect of clear thinking, and this has to involve Bayesian analysis and using maximum entropy to understand what we know before we look at the evidence and how the evidence modifies what we know [and I’m not saying these are easy tools to use]. But this doesn’t get us very far in understanding real world economic and environmental problems.

Maybe not as far as we’d like, but they’re essential tools. In “week303” (which doesn’t exist yet — this link will work someday but not today), I’ll start talking with Nathan Urban about his use of Bayesian methods in economic and environmental problems.

I’d be rash to comment but I feel that the place to look has to be flows in configuration space, and the principle of Maximum Entropy Production will be the key to understanding that.

I’m not sure, but I do want to get to the bottom of the principle of Maximum Entropy Production. I recommend that everyone read this paper:

• E. T. Jaynes, Macroscopic prediction, in H. Haken (ed.) Complex systems – operational approaches in neurobiology, Springer-Verlag, Berlin, 1985, pp. 254–269.

and ask me questions! If you ask questions, that’ll prod me into thinking more about this stuff.

Does anyone out there know if the approach outlined in the Jaynes paper has since led to subsequent significant progress or even breakthroughs in the study of complex systems?

I hope some experts chime in here. My impression, recorded here, is that instead of making ‘breakthroughs’, people are still struggling through the thicket of ideas surrounding the principle of maximum entropy production, or what Jaynes called the ‘principle of maximum caliber’.

(On the other hand, his ‘principle of maximum entropy’ has become quite important as a method for statistical inference, though it still kicks up controversy.)

No expert, but maximum entropy routines are, well, *routine*. Essentially every numerical modeling system appropriate for complex systems analyses implements a variety of maximum entropy methods. Jaynes was a maxent harbinger and strong proponent, but not first or only.

To repeat myself in a different way, to always use maximum entropy (nee maximum likelihood) approaches commits the same error as always using greedy algorithms (e.g. for the vertex cover problem). And for the same reasons, even though they are provably suboptimal, in the “real world” both maxent and greedy approaches are used every day because they work.

No expert, but maximum entropy routines are, well, *routine*.

It’s nice to hear more about this. Thanks!

But still — not to be snarky, just painfully clear — I hope everyone realizes that I wasn’t asking for experts to comment on maximum entropy methods in statistical inference. I mentioned that stuff only to emphasize that it’s not what Charlie Clingen and I are wondering about. We’re wondering about Jaynes’ ‘principle of maximum caliber’ and the closely related ‘principle of maximum entropy production’, which are supposed to shed light on nonequilibrium statistical mechanics. Clingen and I are wondering what progress, if any, has been made with the help of these ideas.

Although presumably my quick scan method for responding to keywords is flawed, I’m not sure what precision is required to discern between the remark that maxent is a routinely implemented method for complex systems analysis and some requested remark on whether maxent principles have been used to shed light on complex systens.

Although presumably my quick scan method for responding to keywords is flawed, I’m not sure what precision is required to discern between the remark that maxent is a routinely implemented method for complex systems analysis and some requested remark on whether maxent principles have been used to shed light on complex systems.

Sorry — I could be confused, but I was trying to distinguish between:

1) the quite routine use of ‘MaxEnt methods’ in statistical inference — where you roughly try to choose the ‘least committal’ hypothesis that fits the given data

It’s the latter that Robert and Charlie and I are wondering about.

Now, I won’t be surprised to discover that these are ultimately just two faces of the same coin. But still, right now, 1) seems fairly clear to me, while 2) seems shrouded in mystery — especially because Prigogine has successfully used a principle of least entropy production in his work on nonequilibrium thermodynamics!

Now, thanks to your remark, I just found a paper that claims ‘MaxEnt’ and the ‘Principle of Maximum Entropy Production’ are two faces of the same coin:

Abstract: Jaynes’ maximum entropy (MaxEnt) principle was recently used to give a conditional, local derivation of the “maximum entropy production” (MEP) principle, which states that a flow system with fixed flow(s) or gradient(s) will converge to a steady state of maximum production of thermodynamic entropy (R.K. Niven, Phys. Rev. E, in press). The analysis provides a steady state analog of the MaxEnt formulation of equilibrium thermodynamics, applicable to many complex flow systems at steady state. The present study examines the classification of physical systems, with emphasis on the choice of constraints in MaxEnt. The discussion clarifies the distinction between equilibrium, fluid flow, source/sink, flow/reactive and other systems, leading into an appraisal of the application of MaxEnt to steady state flow and reactive systems.

And he even has references listing applications to climate change. So, I need to read this and try to understand it.

(Btw, my descriptions of ‘MaxEnt methods’ and the ‘Principle of Maximum Entropy Production’ were very rough — just enough to distinguish them, I hope.)

Right. maxent (without Capitals, if that’s warmer and fuzzier) methods are not just for static inference. They are routinely used for example in *predictive* risk software, not just for physical systems, and as dropdown menu items in commercial operations research software. Entropy of data is there a selectable function, among others that often work better, to *incrementally* maximize.

A little more of my background. As a sometime student of Norman Packard, as a complex systems class project in 1988 he gave me some dozens of gigabytes of meteorology data, of which the data of interest involved the global dynamics of ozone. One of several predictive analyses I did was to maximize the change in an information measure from timepoint to timepoint, copying algorithms from a student of Jaynes. Another analysis that also did not work at all well used a Kalman filter I wrote, but I did not write a good Kalman filter until 1993. Anyway only ad hoc time series analysis worked well. He may remember that I used his account to overutilize the supercomputer facilities. The reason for saying all that was that over 20 years ago maximum entropy production (again, maybe without CapiTals) was well known enough to be suggested as a classroom application for complex systems analysis.

Although I guess this discussion is drifting further afield, it is worth mentioning than implementations of statistical inference techniques based on Kullback-Leibler treatment of priors is most widely distributed in commercial software for image processing, such as ERDAS IMAGINE commonly used in ecological studies. For these purposes people generally use variations such the Transformed Divergence because they work better. I find that I can manipulate distributions (e.g. enforce convergence) much better using a Jensen-Shannon metric than other measures.

FWIW, i.e. 0, it is possible to get an unemergent unmicroscopic version of entropic dynamics of metrics, with square roots of determinants etc., having nothing to do with coordinates.

Jaynes is pretty good. Let me preface my remarks by pointing out the previous sentence as some evidence that I’m not wholly biased against him. I had read him prior to hearing him talk some half-dozen times, and he was always imperious, misogynistic, and 100% dismissive of all questions and criticism. In addition he not only wanted to be the 20th century Gibbs, he thought he was.

He got a lot (a lot!) of applied research money to analyze other people’s data using maximum entropy methods. Many of those results involve what I will call “one shot” data. From page 7 of the linked paper he avers “In geophysics or economics it is seldom possible to repeat an experiment at all.”

I’m going to claim here that failing to properly hedge your bets, i.e. to always bet on the maximum, is always eventually a failing strategy. Also, that the nature of probability is such that you can’t expect maximum to always hold.

It is a pity you predated with the wiki creation the MO birthday by one day :) So the birthdays are Sep 27 for azimuth wiki; Sep 28 for MathOverflow and Oct 28 for nlab.

I’ve got the impression that the Forum is working very slowly (almost endless “loading” and the preview doesn’t seem to react…) in addition to the fact that the Wiki appears down (but they were on different servers, right?)

You can use Markdown or HTML in your comments. You can also use LaTeX, like this: $latex E = m c^2 $. The word 'latex' comes right after the first dollar sign, with a space after it. Cancel reply

You need the word 'latex' right after the first dollar sign, and it needs a space after it. Double dollar signs don't work, and other limitations apply, some described here. You can't preview comments here, but I'm happy to fix errors.

Yay!

Now before people go wild creating new pages, please have a look at the forum discussion on conventions:

http://www.math.ntnu.no/~stacey/Mathforge/Azimuth/comments.php?DiscussionID=7

John has decided to go with Wikipedia’s convention regarding capitalization of page names:

http://en.wikipedia.org/wiki/Wikipedia:Naming_conventions_%28capitalization%29

You’ll save the Azimuth elves a lot of trouble if you could please comply with conventions when creating new pages.

Be kind to your Azimuth elves!

I will probably post more on ‘how to use the Azimuth Project’ — and everyone is encouraged to ask questions here — but also, people should check out the How to page.

Great! (I’m in a hurry, have not yet worked with any wiki, etc., and will be mostly gone for this week… so: ) An Azimuth Project entry I (or you) will (should) be doing is “Carbon Negative Energy”. Got some links on my blog, could be more. The only source of carbon negative energy currently known is

wood gas leaving behind char coal. John knows about it. But many engineers seem to disregard it. Being too primitive the idea seems to insult the tech ego. Even the “sustainable energy” business seems not to be interested. They prefer to sell solar and biogas, which, being not carbon negative is actually not sustainble – It’s a baby step in the right direction taken decades too late… That’s why I couldn’t wait till next week to mention this.I created an entry:

• Carbon negative energy.

It was incredibly easy to do this. I just typed “Carbon negative energy” into the search box on top, and since a page by the name did not yet exist, I was given the option to create one with a single click of the mouse.

Then a place to type stuff appeared, and I typed some stuff in there. Nothing very interesting yet! Just the basic idea.

Then I clicked the “Submit” button at the bottom, and the page was created.

I hope you add more, Flori. To do so, just click on the link above, then click “Edit” at lower left. Then type in stuff.

Don’t worry too much— it’s almost impossible to do anything that can’t easily be fixed.The most important thing is to enter useful information. The second most important thing is to write clearly. Formatting and technical issues come last: the lab elves can clean up any mess you leave. So don’t use those issues as an excuse to put off saving the planet!

Of course,

afteryou get into the habit of entering information in Azimuth, it’s good to learn some tricks for doing it well. For that, start here:• How to get started

and for more detail, try:

• How to

Or, ask questions here!

The article on Carbon negative energy is just a pathetic stub, but I just created two slightly more substantial articles on Florifulgurator’s favorite form of carbon negative energy:

• Biochar

• Terra preta

Check them out and add information and questions!

I won’t be able to create every article that someone wants — you folks have to create them yourselves! But biochar and terra preta are incredibly important, potentially, and I happened to have some information about them at hand…

Paul Krugman has some relevant things to say in his blog today (http://krugman.blogs.nytimes.com/2010/09/27/the-power-of-conventional-wisdom/):

Then lower down:

Hmm. On the other hand we need genuine subject-matter expertise, which Krugman provides on Economics. On the other hand it is very easy for experts to be too narrowly focussed. I keep asking questions on his blog along the lines of “If world-wide demand were raised by world-wide government action to the level necessary to get unemployment down, then how much oil production would the economy need, and can we produce that much oil at the moment?”. My point is that (a) we need government stimulus directed to produce cheap energy to be an (imperfect) substitute for oil; and (b) we need to also soak up private demand (energy crisis bonds?) to prevent smashing (again) into the limit of oil production and crashing the economy. But am I right? As Krugman and others point out, things are very similar to Japan 15 years ago when there was no energy or other shortage. We are missing any way to think clearly about these things. Just different groups of people using lots of adjectives and doing lots of arm-waving (yes there are lots of numbers being thrown around, but their implications are far from clear).

Which brings us back to the question of who has a special monopoly on wisdom. I claim that the subject matter of mathematics is how to think clearly about problems. Yes mathematicians spend time thinking about unimportant problems that they just happen to be able to describe succinctly. And also trying to understand mathematics itself better. But the real problems that drive mathematics are in the real world. Inference is a universal aspect of clear thinking, and this has to involve Bayesian analysis and using maximum entropy to understand what we know before we look at the evidence and how the evidence modifies what we know [and I’m not saying these are easy tools to use]. But this doesn’t get us very far in understanding real world economic and environmental problems. I’d be rash to comment but I feel that the place to look has to be flows in configuration space, and the principle of Maximum Entropy Production will be the key to understanding that.

Robert wrote:

Back when I was a pure mathematician I would complain about people using the world ‘real’ so often like this. Mathematics has its own reality with its own real problems which are beautiful and profound. Unfortunately us rats squirming in the muck of a suffering planet have our own urgent problems, and I’ve decided I need to think about those. These problems aren’t more ‘real’, but we’re stuck with them.

Maybe not as far as we’d like, but they’re essential tools. In “week303” (which doesn’t exist yet — this link will work someday but not today), I’ll start talking with Nathan Urban about his use of Bayesian methods in economic and environmental problems.

I’m not sure, but I

dowant to get to the bottom of the principle of Maximum Entropy Production. I recommend that everyone read this paper:• E. T. Jaynes, Macroscopic prediction, in H. Haken (ed.)

Complex systems – operational approaches in neurobiology, Springer-Verlag, Berlin, 1985, pp. 254–269.and ask me questions! If you ask questions, that’ll prod me into thinking more about this stuff.

The other papers listed here might also be good.

JB said:

Is the link broken?

Fixed, thanks.

Does anyone out there know if the approach outlined in the Jaynes paper has since led to subsequent significant progress or even breakthroughs in the study of complex systems?

I hope some experts chime in here. My impression, recorded here, is that instead of making ‘breakthroughs’, people are still struggling through the thicket of ideas surrounding the principle of maximum entropy production, or what Jaynes called the ‘principle of maximum caliber’.

(On the other hand, his ‘principle of maximum entropy’ has become quite important as a method for statistical inference, though it still kicks up controversy.)

No expert, but maximum entropy routines are, well, *routine*. Essentially every numerical modeling system appropriate for complex systems analyses implements a variety of maximum entropy methods. Jaynes was a maxent harbinger and strong proponent, but not first or only.

To repeat myself in a different way, to always use maximum entropy (nee maximum likelihood) approaches commits the same error as always using greedy algorithms (e.g. for the vertex cover problem). And for the same reasons, even though they are provably suboptimal, in the “real world” both maxent and greedy approaches are used every day because they work.

John F. wrote:

It’s nice to hear more about this. Thanks!

But still — not to be snarky, just painfully clear — I hope everyone realizes that I

wasn’tasking for experts to comment on maximum entropy methods in statistical inference. I mentioned that stuff only to emphasize that it’snotwhat Charlie Clingen and I are wondering about. We’re wondering about Jaynes’ ‘principle of maximum caliber’ and the closely related ‘principle of maximum entropy production’, which are supposed to shed light on nonequilibrium statistical mechanics. Clingen and I are wondering what progress, if any, has been made with the help of these ideas.Although presumably my quick scan method for responding to keywords is flawed, I’m not sure what precision is required to discern between the remark that maxent is a routinely implemented method for complex systems analysis and some requested remark on whether maxent principles have been used to shed light on complex systens.

If you are focusing on dynamics for climate, for instance

http://www.sciencemag.org/cgi/content/full/299/5608/837

is the first Google hit. Maxent works ok. Just isn’t necessarily the best, etc.

John F. wrote:

Sorry — I could be confused, but I was trying to distinguish between:

1) the quite routine use of ‘MaxEnt methods’ in statistical inference — where you roughly try to choose the ‘least committal’ hypothesis that fits the given data

and

2) the apparently more obscure ‘Principle of Maximum Entropy Production’ or ‘Principle of Maximum Caliber’ in nonequilibrium thermodynamics, which says roughly that a physical system will do whatever it can to produce as much entropy as possible.

It’s the latter that Robert and Charlie and I are wondering about.

Now, I won’t be surprised to discover that these are ultimately just two faces of the same coin. But still, right now, 1) seems fairly clear to me, while 2) seems shrouded in mystery — especially because Prigogine has successfully used a principle of

leastentropy production in his work on nonequilibrium thermodynamics!Now, thanks to your remark, I just found a paper that claims ‘MaxEnt’ and the ‘Principle of Maximum Entropy Production’

two faces of the same coin:are• Robert Niven, Jaynes’ MaxEnt, Steady State Flow Systems and the Maximum Entropy Production Principle.

And he even has references listing applications to climate change. So, I need to read this and try to understand it.

(Btw, my descriptions of ‘MaxEnt methods’ and the ‘Principle of Maximum Entropy Production’ were very rough — just enough to distinguish them, I hope.)

Right. maxent (without Capitals, if that’s warmer and fuzzier) methods are not just for static inference. They are routinely used for example in *predictive* risk software, not just for physical systems, and as dropdown menu items in commercial operations research software. Entropy of data is there a selectable function, among others that often work better, to *incrementally* maximize.

A little more of my background. As a sometime student of Norman Packard, as a complex systems class project in 1988 he gave me some dozens of gigabytes of meteorology data, of which the data of interest involved the global dynamics of ozone. One of several predictive analyses I did was to maximize the change in an information measure from timepoint to timepoint, copying algorithms from a student of Jaynes. Another analysis that also did not work at all well used a Kalman filter I wrote, but I did not write a good Kalman filter until 1993. Anyway only ad hoc time series analysis worked well. He may remember that I used his account to overutilize the supercomputer facilities. The reason for saying all that was that over 20 years ago maximum entropy production (again, maybe without CapiTals) was well known enough to be suggested as a classroom application for complex systems analysis.

Although I guess this discussion is drifting further afield, it is worth mentioning than implementations of statistical inference techniques based on Kullback-Leibler treatment of priors is most widely distributed in commercial software for image processing, such as ERDAS IMAGINE commonly used in ecological studies. For these purposes people generally use variations such the Transformed Divergence because they work better. I find that I can manipulate distributions (e.g. enforce convergence) much better using a Jensen-Shannon metric than other measures.

FWIW, i.e. 0, it is possible to get an unemergent unmicroscopic version of entropic dynamics of metrics, with square roots of determinants etc., having nothing to do with coordinates.

Jaynes is pretty good. Let me preface my remarks by pointing out the previous sentence as some evidence that I’m not wholly biased against him. I had read him prior to hearing him talk some half-dozen times, and he was always imperious, misogynistic, and 100% dismissive of all questions and criticism. In addition he not only wanted to be the 20th century Gibbs, he thought he was.

He got a lot (a lot!) of applied research money to analyze other people’s data using maximum entropy methods. Many of those results involve what I will call “one shot” data. From page 7 of the linked paper he avers “In geophysics or economics it is seldom possible to repeat an experiment at all.”

I’m going to claim here that failing to properly hedge your bets, i.e. to always bet on the maximum, is always eventually a failing strategy. Also, that the nature of probability is such that you can’t expect maximum to always hold.

It is a pity you predated with the wiki creation the MO birthday by one day :) So the birthdays are Sep 27 for azimuth wiki; Sep 28 for MathOverflow and Oct 28 for nlab.

You should have told me sooner!

I’ve got the impression that the Forum is working very slowly (almost endless “loading” and the preview doesn’t seem to react…) in addition to the fact that the Wiki appears down (but they were on different servers, right?)

They’re on different servers. I’ll email Andrew Stacey in case he doesn’t see my comment on the Forum. I’ll cc you, so you know his email address.