This might come in handy for someone having trouble logging into BTFON (Public WiFi hotspots).

The BT broadband base units have the ability to support public WiFi (as of 2009 it's enabled by default - yeh you should have read the small print eh). It available as BTFON (I think it used to be called OpenZone?).

Anyway as a BT broadband customer I get to use these for free or else you can pay for access [clock] time on the spot. It took me a while to figure out how to log in.

The three options are, use a username/password from:

- BT Total broadband
- BT OpenZone
- BTFON

I have a BT username/address ending with @btopenworld.com (my DSL login username which is supposed to be the username I use to access BTFON). However this did not work on either of the three options (I wasn't actually 100% sure which of the three options to choose).

In the end I logged in using option 1 (BT Total broadband), and changing the end of the email/username to:

@btinternet.com

The password is the same as your DSL login password (which I had to reset as it happens as the one I had written down didn't work!). And it took about a minute to actually login.

Instant access my arse.

BTFON hotspot map

1) Random forests (RF) is(are) a competitive data modelling/mining method.

2) An RF model has one output - the output/prediction variable.

3) The naive approach to modelling multiple outputs with RFs would be to construct RF for each output variable. So we have N independent models, and where there is correlation between output variables we will have redundant/duplicate model structure. This could be very wasteful indeed. Also as a general rule more model variables implies a more overfit model (less generalisation). Not sure if this applies here but it probably does.

In principle we could have an RF with multiple outputs. The prediction variable is now a vector (n-tuple). The decision nodes in each decision tree are now splitting the set of target/prediction vectors based on a threshold vector, I figure this threshold is taken to be a plane in the n-dimension space and that therefore we can determine which side of the threshold vector each of the target vectors is on.

The optimal prediction value for each side of the decision split is the mean of the all of the vectors on that side (what sort of mean?).

Finding the optimal split point when working with single variables is trivial and computationally fast/efficient. For an n-tuple we cannot find the optimal split (or at least it becomes computationally infeasible as N increases), but we may be able to find a near optimal split using a monte carlo type method (or some hybrid of monte carlo and local gradient traversal).

Would this actually work? That is, would it just map the training pairs without generalising?

What techniques would such an approach be similar too?

I have reasonably high confidence that I'm homing (honing?) in on the root problem...

Hypomagnesemia
Also, hypomagnesemia is related to thiamine deficiency because magnesium is needed for transforming thiamine into thiamine pyrophosphate.

Thiamine
Magnesium, which is required for the binding of thiamine to thiamine-using enzymes within the cell, is also deficient due to chronic alcohol consumption. The inefficient utilization of any thiamine that does reach the cells will further exacerbate the thiamine deficiency.

Mg citrate found more bioavailable than other Mg preparations in a randomised, double-blind study
bioavailability of three preparations of Mg (amino-acid chelate, citrate and oxide) were compared at a daily dose of 300 mg of elemental Mg in 46 healthy individuals.
...
We conclude that a daily supplementation with Mg citrate shows superior bioavailability after 60 days of treatment when compared with other treatments studied

Magnesium bioavailability from magnesium citrate and magnesium oxide
[For magnesium citrate] The increment in urinary magnesium following magnesium citrate load (25 mmol) was significantly higher than that obtained from magnesium oxide load (during 4 hours post-load, 0.22 vs 0.006 mg/mg creatinine, p less than 0.05; during second 2 hours post-load, 0.035 vs 0.008 mg/mg creatinine, p less than 0.05). Thus, magnesium citrate was more soluble and bioavailable than magnesium oxide.

Magnesium deficiency (medicine)
Magnesium is absorbed orally at about 30% bioavailability from any water soluble salt, such as magnesium chloride or magnesium citrate. The citrate is the least expensive soluble (high bioavailability) oral magnesium salt available in supplements, with 100 mg and 200 mg magnesium typically contained per capsule or tablet.

Magnesium aspartate, chloride, lactate, citrate and glycinate each have bioavailability 4 times greater than the oxide form and are equivalent to each other per amount of magnesium, though not in price.[6][7]

mproving Magnesium Absorption and Bioavailability
Magnesium Salts
[...]
In contrast, another study focused on the effects of magnesium supplementation in 40 elderly magnesium-deficient patients and compared oral versus intravenous administration (Gullestad et al., 1991-1992). The oral magnesium lactate-citrate preparation was given for six weeks at a daily dose of 15 mmol; the IV magnesium sulfate formulation was given at a daily dose of 30 mmol as an infusion in 1000 mL of saline for seven days. The two routes of magnesium administration yielded comparable results. The authors termed bioavailability of oral magnesium lactate citrate "satisfactory" and concluded that oral delivery of magnesium supplements for six weeks may restore magnesium levels in magnesium-deficient patients.

Bioavailability of US commercial magnesium preparations
Results indicated relatively poor bioavailability of magnesium oxide (fractional absorption 4 per cent) but significantly higher and equivalent bioavailability of magnesium chloride, magnesium lactate and magnesium aspartate. We conclude that there is relatively poor bioavailability of magnesium oxide, but greater and equivalent bioavailability of magnesium chloride, lactate, and aspartate.

Very interesting stuff...

Gamma-Linolenic Acid (GLA)

I've been taking a small amount for a few weeks and have started increasing the dose as it has greatly improved my complexion (along with 2*25mg B6/day which I've been taking for, I dunno, a couple of years or so now) to the extent that this is now effectively a solved problem. Will report back later.

Possibly B6 is implicated in synthesis of GLA...

...co-factors that are essential for the metabolism of EFAs so that adequate amounts of PUFAs could be formed endogenously such as low-dose aspirin, folic acid, vitamin B6, B2, L-arginine, vitamin C and Mg2+...

Also moderately interesting was a random email to me suggesting that people of Celtic origin are typically deficient in GLA. Need to look into that one.

======
Gamma Linolenic Acid: A Natural Anti-inflammatory Agent—Part I

NUTRITIONAL AND MEDICAL IMPORTANCE OF GAMMA-LINOLENIC ACID