Log in

Surfing the Metaverse
Recent Entries 
Just wondering if this has been tried.

I.e. instead of regularizing by decaying factors towards zero, randomly drop-out half of the data points on each training pass (as per drop-out in neural nets); the other factors must then adjust to make up for the missing data. When multiplying the side vectors to produce the modeled matrix, we just half the contribution from each pair of side vectors (R pairs for rank R).

As per drop-out in neural nets, this simulates a massive ensemble with shared weights between the ensemble models.
27th-Oct-2016 12:19 am - If I have seen further...
"If I have seen further, it's because I was looking"

Is perhaps a more faithful version of Newton's comment. Most people just aren't looking, in fact most people are actively avoiding looking.
25th-Oct-2016 11:54 pm - Bitcoin tx fees
Keep an eye on...


Once block capacity was reached a market formed defined by varying demand for transaction versus a fixed supply. Hence the change in total fees per unit time is an indication of demand.
Where Will All the Money Go When All Three Market Bubbles Pop?

It feels like the early to mid 2000's all over again. Back then we had inverted yield spreads, i.e. the risk premium wasn't just flat but inverted in some cases, and across all major asset classes everywhere. Essentially there's too much money chasing too little actual assets and wealth, which is just another way of saying that the money is overvalued.

Consider defining a new measure of value based on quality of life (let's call it a Qual) and the ability of certain assets to provide quality life; so things like food, housing, infrastructure, hospitals, schools and access to a high quality labour market.

Qual is just a measure, so we can say that a town can be said to have k quals of quality on it, and if we had two such towns there are twice as many quals. So as an economy expands and contracts the absolute number of quals rises and falls. And what a qual is worth in terms of real life stuff remains constant, so you could define a price in quals for food, housing, fuel, building materials and that price would remain constant, since a housing a food, etc. provide the same utility today as they did 50, 100, 1000 years ago. And yet the price in dollars of some of these things is high, which really means that the value of dollars is falling, measured in quals.

I.e. if the price of everything goes up and is in bubble territory, then your're not really looking at an all-asset price bubble, but rather monetary inflation. And yet the official price inflation statistics suggest inflation is low to modest, even near zero. This may be one of those stupid situation where post financial crash everyone says that, well of course in hindsight we were staring at a massive problem all along, but we never consider warnings signs X Y and Z in those terms... yadda yadda.

Essentially inflation is being measured by a peculiarly constrained subset of assets ('basket of goods') that seems to avoid all of the stuff experiencing asset bubbles (for starters consider the difference between 'goods' and 'assets'). The price of consumer goods are subject to lots of dynamics and price pressures, and I would say they don't really capture what's going on with global money systems; and yet these inflation stats are quoted as evidence of low inflation.
Interested in improvements on this attempt:

Intuitive explanation of entropy?
10th-Oct-2016 04:41 pm - Zcash

A bitcoin derivative that appears to be sufficiently distinct from bitcoin to be interesting, i.e. considerably more so than litecoin or dogecoin.

I'm guessing it could take a chunk out of bitcoin's 'market capitalisation' within a short period. One to watch closely for the time being.

Installation instructions (for linux) (all went smoothly for me):


To mine you just need to add 'gen=1' to the config file. A bit more info here:

9th-Oct-2016 07:23 pm - Estimating a Biased Coin
Some foundational maths towards some AI related work I'm looking into:

Estimating a Biased Coin

Feedback welcome (maths, mathjax layout, language/terminology, typos...)
This page was loaded Dec 8th 2016, 3:57 am GMT.