Can We Agree On How To Measure The Similarity Of Dive Profiles?

[dmaziuk]

One of the debates is actively going on in this thread, and it currently involves trying to determine whether two ascents are similar.

From the look of the graphs, the profiles posted are obviously described by the same equation with slightly different coefficients. So mathematically they're similar. The last couple of profiles have a difference of 30 minutes so they're very different in most other respects.

 

[stuartv]

Why not put a stake in the ground? Why not propose a definition for "similar"?

Here's one:

Use Buhlmann ZH-L16B (no GF) to calculate the tissue tensions in every compartment for every 1 minute interval of each of the profiles in question. Compare all the tissue tensions. If none of them are ever different by more than 10% at any point in the dive, then the profiles are "similar".

 

[boulderjohn]

You still seem to think there are definitions for similar and different - they don't exist.

A number of years ago, a prolific PADI-basher on ScubaBoard said he can speak with authority on PADI and its program because he had "recently" been a PADI instructor. In fact, he used the word "recent" to describe his experience several times in several posts and several threads. Knowing the details, I challenged him, saying (at that time) he had not been a PADI member for 26 years. His membership had ended, in fact, before PADI had dramatically changed its instructional approach, so he had no experience with the then current curriculum whatsoever. In my wording, I strongly implied that he was being dishonest in the hope that people would be fooled into believing he had knowledge that he did not in fact have.

Not so, he replied. In his opinion, 26 years was quite recent, so he was not misusing the term at all and thus was not being deceptive with its use..

Well, in geologic terms, 25 years is indeed extremely "recent." The tectonic plates have barely moved in that time. In the context of that discussion, though, most people would consider a quarter century to be a very long time, not remotely recent. Although most people would hold those views, apparently not all people, so the word "recent" could not be accurately defined. It can only be judged by individuals in the context in which it is used.

It is the same for words like "similar" and "different." The words cannot be strictly defined and can only be judged in context. Moreover, in order to judge them in context, you have to understand the context. In diving, people who are not familiar with the kind of diving being discussed in that thread will have trouble understanding the context.

Let's consider two dive profiles with only a 5 minute difference in bottom time to show what I mean.

1. The first is a recreational dive to 60 feet for 30 minutes. Add 5 minutes to it, and the difference is not worth talking about. In both cases, you will make a normal ascent, do a safety stop, and head for the surface. The difference in the amount of gas used is tiny, and even the worst air hog should be able to go just fine on an AL 80. I think just about everyone would say those dives are similar.

2. The second is a technical dive to 200 feet for 30 minutes. Add 5 minutes to it, and the difference is huge. I just ran two profiles through a desktop computer program, and that extra 5 minutes of bottom time adds more than 15 minutes of decompression to the dive, for more than 20 minutes more time under water. The total expected gas usage for the longer dive is more than 40 more cubic feet, which will impact the amount of gas that needs to be carried. I think that just about everyone would say those extra 5 minutes are moving the dive profiles over to the "different" category.​

 

[dmaziuk]

Why not put a stake in the ground? Why not propose a definition for "similar"?

Here's one:

Use Buhlmann ZH-L16B (no GF) to calculate the tissue tensions in every compartment for every 1 minute interval of each of the profiles in question. Compare all the tissue tensions. If none of them are ever different by more than 10% at any point in the dive, then the profiles are "similar".

OT: I find it amusing that you wouldn't criticize the scubalab "study" for neglecting to do this exact thing before they published their "algorithm performance results".

 

[agilis]

The purposes of comparison are critical elements in establishing any protocol which involves 'similar' profiles. In addition to obvious variables like temperature and physical effort resulting from water movement there is that most elusive but vital variable, the specific human involved. The variable responses and reactions of different divers should be so obvious as to require no explanation, but even for the same diver changes over time, even over very short periods, makes precise equivalency difficult to establish.

Even minor variations in ambient conditions will compromise any equivalency measurement, but the addition of the human factor makes precision close to impossible, except theoretically.

 

[stuartv]

OT: I find it amusing that you wouldn't criticize the scubalab "study" for neglecting to do this exact thing before they published their "algorithm performance results".

A classification of "similar" implies an evaluation of the data. My previous post was just an off-the-cuff first draft of a way to evaluate dive profile data to make a determination of "similar" or "not similar".

The SL reports provided the actual data, leaving the reader to evaluate it based on their own needs, priorities, diving, etc.. The SL reports are the equivalent of providing 2 dive profiles and letting each person decide for themselves whether they qualify as "similar" to that person.

I don't find that to be something to criticize. I find it to be laudable. It is in keeping with my constant theme of "provide the people the objective information and let them form their own judgments." This is in contrast to a constant SB theme of evaluate, judge, only pass along the judgments, then criticize people who question them or make different judgments.

 

[dmaziuk]

Even minor variations in ambient conditions will compromise any equivalency measurement, but the addition of the human factor makes precision close to impossible, except theoretically.

But if deco models are, for all practical purposes, theoretical, then if the purpose is to compare those models the human factor shouldn't be added. One could argue that it should, in fact, be eliminated. Numbers are so much easier without those pesky humans.

 

[kr2y5]

Well, in geologic terms, 25 years is indeed extremely "recent." (...) It is the same for words like "similar" and "different."

It's not the same. Note that in your own example, you are referring to a standard framework for expressing the distance in time. Rather than relying on a subjective judgment, where all one can deduce on a case by case basis is "recent" or "not recent", we have the ability to express the distance in time as a number, in a manner that is commonly accepted. Rather than saying "recent", which is purely subjective, we gained the ability to say "25 years ago", which elevates it from subjective to objective. While context is important, not every bit of context carries the same significance. You don't need to know that during those 25 years, someone got married, had kids, and a foreclosure.

"similar" and "different." The words cannot be strictly defined and can only be judged in context.

As I think I made it clear in earlier posts, including the OP, I'm not looking for a golden formula that provides a binary decision "similar or not" for any set of circumstances. I'm looking for ideas how to quantify similarity in a way that is practically useful.

One can certainly come up with metrics of similarity that are mathematically well-defined, such as the maximum difference in time spent at any given depth, that particular one just seems very limited in a number of ways, including the ones you pointed out.

Earlier in this thread, Richard suggested there might be ways to define similarity in a manner that factors in physiology, risk, and other aspects. Personally, I'd be interested in hearing other ideas for how that could be done, what are their benefits and shortcomings.

 

[agilis]

But if deco models are, for all practical purposes, theoretical, then if the purpose is to compare those models the human factor shouldn't be added. One could argue that it should, in fact, be eliminated. Numbers are so much easier without those pesky humans.

Deco computers are actually making an estimate of one and only one phenomenon, the persistence of nitrogen gas in tissue, and are basing this estimate on time/depth permutations established primarily through experience and physiological analysis.

There are a number of models employed in this mechanistic non-organic measurement, models which are generally described as being on a spectrum ranging from conservative to liberal.

None are 100% reliable because they do not measure anything actually happening to the diver using the computer. They provide purely theoretical information, and would, under various circumstances, provide reassuring information to a dead diver.

Metrics of simlarity are fairly easy to establish for computers or other mechanical object. For humans the variations are endless.

 

[dmaziuk]

Deco computers are actually making an estimate of one and only one phenomenon, the persistence of nitrogen gas in tissue, and are basing this estimate on time/depth permutations established primarily through experience and physiological analysis.

Pressure, strictly speaking: depth as shorthand for pressure.

One could calculate gas loading at, say, 2-minute resolution for the entire dive, treat this as a sequence of values, then run a sequence similarity search on two of those. Commonly used sequence similarity algorithms include phoneme matching in natural language processing (speech recognition) and residue matching in DNA/RNA/protein sequences in bioinformatics.