algorithm comparison app?

[baggins_69]

I'd be interested in comparing the different algorithms against historical dive data.

are the various algorithms documented/available? if so, can you point me to them?

I'm a software engineer and thinking it would be fun to write a quick application that would run the same recorded data against multiple algorithms and graphing various aspects (nitrogen loading, NDL, yadda yadda)

 

[grantmac]

Multideco can run Buhlman or VBM.

But honestly there is a reason almost every serious diver uses Buhlmann now, so just get Subsurface (or use decobuddy online) for free and start looking at what matters which is the dive profile and gas selection.

 

[tursiops]

are the various algorithms documented/available?

Only a few are available. Most are proprietary (i.e., Oceanic, Aqualung, ScubaPro, Suunto, MARES, Cressi, etc). Even when one of the manufactures uses an open algorithm, you don't quite the same result becasue of various internal decisions, even though the main algorithm is the same. Shearwater and SubSurface both use Buhlmann, for example, but do not give the same results. And MultiDeco's Buhlmann is different also.

 

[scubadada]

It is not a simple matter to compare decompression algorithms. The first, clean dive NDL is one thing, how they deal with repetitive dives is yet another.

 

[Mobulai]

I'd be interested in comparing the different algorithms against historical dive data.

are the various algorithms documented/available? if so, can you point me to them?

I'm a software engineer and thinking it would be fun to write a quick application that would run the same recorded data against multiple algorithms and graphing various aspects (nitrogen loading, NDL, yadda yadda)

Here is a Buhlmann lib that can save you some time and not starting from scratch, but the dive logs format doesn’t match anything I’ve seen around; you’ll have to mess around with the structures or make new dive-log structs

Speaking of, what „historical data“ do you plan to run there? Is there an „open pool“ like kitty-data? I only heard of NDEU tests but these aren’t much, I am looking too so hints are welcome

I don’t know where VPM or RGBM or DSAT docs/implementations are at, but I also don’t care that much for them; Bühlmann is quite enough for me.
In case one of the algorithms you wanna „test“ is the proprietary version of Scubapro, the ZHL8/16MB-ADT, there are some „empirical“ „lookup tables“ here, i think also an app of sorts

Pointers on „fair“ algorithm comparison conditions, and why Buhlmann is, well.. the state of the art (awesome lecture by @Dr Simon Mitchell):

Dr Simon Mitchell - Wats is optimal decompression?

Presentation timeline:00:01 - Introduction, Welcome etc (in Polish)3:53 - Simon's presentation starts1:40:29 - Q&A section

www.youtube.com

Subsurface is also OSS so mybe of you can port the heatmap tissue-loading plots would be awesome
———
Edit: Speaking of subsurface, I have a design decision question? Is your goal to replay dives from the past as-is for metrics, or, to look at the deco algo „decisions“ as in NDL/deco stops (and from there look at subsequent loading)?
If you just wanna replay and see tissue loading, use Subsurface directly no need for writing your own. Simply import log and view it graphically.

If you’re looking for option 2… tell us more, I’d love to know how you modify the dives so they become „equivalent“
———
Look forward to seeing what turns up either way

 

[baggins_69]

Speaking of, what „historical data“ do you plan to run there? Is there an „open pool“ like kitty-data? I only heard of NDEU tests but these aren’t much, I am looking too so hints are welcome

I was planning to use the data from the ~120 dives I have downloaded from my current computer.

that gives me a good mix of clean and repetitive dives and I could also tweak the data to shorten or lengthen surface intervals and see how it affects things.

also, subsurface has an 8k dive data set to test with: GitHub - subsurface/large-anonymous-sample-data: Very large divelog file with >8k dives and >4M samples

 

[Mobulai]

that gives me a good mix of clean and repetitive dives and I could also tweak the data to shorten or lengthen surface intervals and see how it affects things.

Subsurface is a great spot to try that for Buhlmann — if you have the patience you can edit your logged dive in planning mode sample by sample, 2nd plot below (no replanning of stops tho of you change GF (or algo if available)— that would be a feature I’d love to see); but would be interesting to see how other algos plot like for start

I think (quite sure actually) you can do that in subsurface desktop version
I was able to import a dive profile from previous logs and you can edit sample points (depth, gas/set point (ccr)…)

In the plot side you can activate the extra overlays for surfGF, TTS, and tissue loading

Lemme get to a pc and get you some screenshots
EDIT:
After you have a dive imported into your log book just click Log > Edit in Planner to use it in the edit mode; or just plan a fresh one Log > Plan Dive

Normal log book view looks like this

View attachment 869867

Planner mode looks like this (dragable sample points in the graph is really helpful)
View attachment 869868

Edit2:
This feature isn’t available on the mobile version of subsurface, Maybe if you wanna still contribute some code, they could use help porting the parts that didn’t (probably impossible w/o completely refactoring, it might be a Herculean task)

Edit3: ignore my whacky (log) example, this is a ccr dive with some setpoint changes that translated wrongly as gas changes, the original log even registers as OC (but the good thing is you can edit it in pos)t; but it’s the only log i tried that feature on (I was looking at tissue loading and surfGF values vs Setpoint choices for variations on that profile)

 

[pisauron]

Multideco can run Buhlman or VBM.

But honestly there is a reason almost every serious diver uses Buhlmann now, so just get Subsurface (or use decobuddy online) for free and start looking at what matters which is the dive profile and gas selection.

What is the reason?

 

[tursiops]

What is the reason?

Trustworthy, efficient, sufficient, affordable, transparent, available...

 

[Mobulai]

^this

What is the reason?

And this

Dr Simon Mitchell - Wats is optimal decompression?

Presentation timeline:00:01 - Introduction, Welcome etc (in Polish)3:53 - Simon's presentation starts1:40:29 - Q&A section

www.youtube.com