thirdwave

Github Mirror

Week 51

Another research, which looked at parliamentary vs presidential systems in terms of economy, came up with a different answer: purely parliamentary system.

Paper by McManus et al.

I dont have the time to dig into this yet (to replicate it in Python) -- here is the data (from Dr. McManus).

Same issue again. Sine waves would be modeling - too much work

eSpeak sucks. Who wrote this shit?

Cloud Speech! Right.

Why is most tts (text-to-speech) code out there based on Deep Shit now?

Speech at its simplest is 1D data. It can be represented by a sum of sine waves - u could extract those base waves through sinusoidal regression. Save those regression results in a DB per each word (how many are there, like 20-30K?) and for TTS, while reading doc per word look up its waves, sum them up, play, repeat.

😡😡😡😡😡😡😡😡😡😡😡😡

💣💣💣💣💣💣💣💥💥💥💥💥

(Of course according to Time for Change model, both Clinton and Bam had a major advantage going into their elections -the incumbent was unpopular due to his delivery-, but, at least they didn't lose like Gore)

In the past 30+ yrs we've seen the opposite (of Corb) from the left leaders - Clinton, Blair, Bama - suave, hipters who can get elected, but cannot deliver left leaving an electoral disaster behind (it is the policy execution that has staying power, not lip biting). Corbyn was stiff who probably could deliver left. Ideally u want both (speak left, be a closer, get elected, then deliver left).

Basically the motto is get elected, somehow, then deliver left, ppl will feel the diff and vote for you. Parties who did not do that have been disappearing - we talk a lot abt US, UK, how Dem equivalents are miserable there, but in Germany too, bcz of "Third Way" Schroder SPD cratered. Clinton, Blair hipsterism will only carry you so far.

#ukelection

"Taiwan has massive amount of motorcycles contributing to high level of particulate matter (PM2.5), & air quality continues to deteriorate. #zeroemission scooter to reduce air pollution"

Link

Hydrogenious LOHC Technologies spoke at Bundestag, on H2

Link

There is no easy way of fighting inflation.

"Volcker raised rates too high during the 80s, he hurt the economy, hurt working people".

[geek] Noticed that didn't you. Very good. Yes, it looks like an objective, a target. Even better that is a convex target which have certain guarantees in opt. So now we can calculate the best weights that'll give the minimum internal correlation.

Hell 1 could use this procedure to filter out unwanted stocks even. Start with many, many symbols and the optimization could zero out some, indicating they are no good, so they are left out [/geek]

"[geek] I noticed the w'H w from optimization - it looks like a objective func [/geek]"

👍

"@GroverNorquist

President Trump is right to fight against expanding the electric vehicle tax credit. Taxpayers shouldn't be on the hook for a subsidy for California’s elite. Thank you @realDonaldTrump for defending American taxpayers"

Link

5700 yrs ago? That's like yesterday. What genius is managing ur science content?

"Stone Age chewing gum holds clues to the life of a young girl who lived 5,700 years ago" -- CNN

"Xi suffered another serious blow in November, when The New York Times obtained more than 400 pages of internal Chinese documents concerning the mass incarceration of ethnic minorities – particularly Muslim Uighurs – in the Xinjiang region. Only Chinese government insiders had access to such sensitive materials, suggesting that Xi’s political enemies may have deliberately leaked them to the Western press in order to undermine his international standing"

Link

MIT prof talking smack on quasi-Newton methods. Dude you are so totally wrong.

What is a low correlation portfolio and why is it good?

It's simple - prices form a time series, then low or negative correlation would mean one is down when the other is not, vica versa. So if one instrument took a hit, the others would save you. The "bad"s cancel out, and, if tests show, overall, the portfolio generates returns, we are a-ok.

Correlation between two series is known to many. How about correlation of 3 series or more? For that, there is the correlation matrix. Say we have Apple, IBM, and Exxon,

import pandas as pd, pandas_datareader.data as web, datetime

start=datetime.datetime(2001, 1, 1)
end=datetime.datetime(2019, 11, 1)
insts1 = ['AAPL','IBM','XOM']
df1 = web.DataReader(insts1, 'yahoo', start, end)['Adj Close']
ret = df1.pct_change()
H = ret.corr()
w = np.ones((1,len(insts1))) * 1./len(insts1)
print ('score', np.float(np.dot(np.dot(w,H), w.T)))

score 0.6021293830308531

Correlation matrix of the prices gives H, then w'H w gives a single score to judge to portfolio by. w is the weights, how we allocate / distribute money to each instrument. For simplicity we picked equal. 1/3, 1/3, 1/3.

In this example we see intra-correlation is high (1 would be highest). Can we do better? There are two tech stocks in there, AAPL and IBM. How about we replace IBM with a retail-ish tech stock? Say Amazon?

insts2 = ['AAPL','AMZN','XOM']
df2 = web.DataReader(insts2, 'yahoo', start, end)['Adj Close']
ret = df2.pct_change()
H = ret.corr()
print ('score', np.float(np.dot(np.dot(w,H), w.T)))

score 0.548369645733854

The correlation went down. If the overall returns prove (backtest) better, this is the better portfolio.

Ideally u want correlation to be a big fat zero. These were just random symbols off the top of my head

The fight. "Learning" statistically as in deep nets vs. formulating the problem somehow starting from principles... A similar version of this fight actually took place between Chomsky and Norvig (of Google) a while back on natural language processing. Chomsky does not like ML methods. I'd say he is right on this one.

(Chomsky knows a thing or two about stats u see, hard sciences and experiment, so can speak on such issues with some authority)

A lot of the successes of the recent AI methods (which are great and I applaud) are in the category of weak AI. This is the first level of pattern recognition - especially for to/fro conversion type tasks if you remain in the same domain, e.g. speech sounds to text of those sounds u can do some damage, the "AI" methods can work. But interpreting that speech, translating its ideas are still far out of reach.

We also talked about clean formulas vs. constants before (which DL ppl call weights or coefficients) in physics context. Basically in that formula vs. weights trade-off, DL chose all weights, all the time. Their "models" such as they are, can have millions of weights. This is the price you pay for not burning brain cells on formulating, modeling a problem.

I've become an expert extracting code from Scipy and flattening it (so everything per feature is in a single, clean file). Any feature, I do it under 1 hour flat.

Sign of incoming trouble - the diff between YoY hiring and wage growth starts to widen. The widening gap is a forward indicator. It makes sense from a macro perspective, in late cycle labor conditions tighten, wages increase, firms' profits start to go down, also their hiring starts to decrease, on and on it goes until the pop

#lateCycle

Ref

Just realized zombie movies hark back to forager-like conditions too (you try to evade, run, hunt, gather), they create these conditions in a modern albeit in a post apocalyptic backdrop. Instead of wild nature u get broken arch (forage 4 food in abandoned supermarket) and ppl who are now monsters (wild animals).

You see, tech, white-collar world is not only about inbread centrist mama's boys. They have a "left" too, and it is in the free, open source crowd.

Removed the GMaps static map api integration 🤮

Offline mapping done; Open source mapsforge allows reading of reprocessed (also free, open) OSM maps. I interface to their data, dump parts I need through their OSS, the rest is offline Py.

U'll be asking for these during the zombie apocalypse my friend. I might give them to u in return for some shotgun shells.

"Company aims to produce world's first fossil fuel-free steel

In the small town of Lulea in northern Sweden, SSAB has teamed up with Vattenfall and LKAB to build a pilot plant for the world's first fossil fuel free steel-making process, which will replace coking coal with hydrogen, and if successful can reduce Sweden's total carbon emissions by 10%"

Link

Realist interpretation of physics can be based on "human interpretable" constructs and still be real - meaning stuff that exists wout observer being there. I think the problem with these comments is its use of "common sense" and "everyday". What is the elements, the clay from which these comments are formed? For the non-technical they are springs, pulleys, balls, wind, elasticity, force, movement, stuff rotating around other stuff (atoms, or planets) or simulacra (easily relatable these days thanks to movies).

The "everyday" items for STEM ppl are graphs, matrices, DEs, the interplay btw the continuous and the discrete.. Graphs can be directed, non-directed, some algorithms can work on them, in various speeds, like shortest path, minimum spanning tree, etc.

This is important, because you can construct a reality from these structures. Like a game - and the root of the game would be a graph. Or a probability distribution (needs pseudorandomness knowledge which isnt really random), still real, same for everyone.

I'd say realists vs crackers are divided half'n half these days, among those who matter... So the former group is nothing to be dismissed.

"There is still something so radical in quantum physics that we cannot translate in our spontaneous everyday ontology and I think all these theories try desperately to translate it back in some kind of understandable for our common sense ontology" -- Zizek

Peak start-up

NYT being a corporate slut. Related.

"Nike’s Fastest Shoes May Give Runners an Even Bigger Advantage Than We Thought"

Link

Trump does the uncouth mixed with (some) style. It is also directed to something, so ppl think "he is tearing shit down for me". It's a tough act though. No wonder his popularity is low: -11% net pop according to Gallup.

They are not people. They are representatives.

The system is unnatural 4 sure. Very few people providing leadership to many on diverse topics is weird. This happens because of modernity which itself is unnatural.

But given that system, representative needs to have more. Say JFK - guy was chasing skirt, humping anything that moved. Does it matter? Well, he was fuckin so many bitches an East German agent slipped in there, now you have a national security problem. Cant do that.

"Does it matter for representatives to be uncouth, crass? They are people after all"

Emilia Clarke looks good in GoT when she doesnt make too many facial expressions. In real life, the actress, when she starts moving those eyebrows, oh-boi. Keep them still girl.

She looks like a nice, lively person tho, dont mean to nag

Obviously the leader, the representatives ppl elect need to be respected (along, ahem, those SNP seats, 48 was it? 😶). It all works out in the long run.

#ukelection

FDR didn't win on the beauty of his "ideas". 1st he almost didnt get the nod but party machine thought he was a loser, easily controllable so they let him. Then he won the general bcz the incumb was obscenely bad. Once in power though, he enacted good policies which ppl felt through their pocketbook, and with a general feeling of gov being on their side and they voted for him, again and again until he died.

First theory still holds - EU and Japan (bcz of their banks / corps) fell into debt deflation territory post 00, w low rates, slow growth.

In US during '08 crisis the FED took the toxic assets off the banks' balance sheet fixing their debt problem so they started lending. EU / Japan did not follow the same course.

"[Paraphr] There are three theories of recent meager growth [and he thinks he debunked the 1st]"

Link

Will we see the formation of an Anglosphere after Brexit? US + UK + AUS plus maybe Japan? All seafaring nations.

I dont think China will be able to play for the big league. If India turns pro US there wont be many other powers left. EU + RU rubs the Eastern states the wrong way, hard to form.

Sun Yat-Sen. Why do u never hear of this guy? It's always Mao, Mao, Mao. This dude gave some thought to democracy in Chinkstan.

"City of London, not to be confused with London is like a city-state within the greater city of London, within England, within the United Kingdom, within the Commonwealth. This city acts as a sort of central location to conduct business with all of these British tax havens in a way that is far too complicated to get into [right now]. [For rich people] London [is a place] to park their money in one of the most established and secure financial centres in the world [and] they also get access to a world-class tax offshoring network that they can conveniently control from the City of London. This explains another anomaly with the economy of the UK it is the world's seventh largest economy by GDP but it attracts the second highest rate of foreign investment in the world"

Link

Tiggers be flailing on IRel - the odd behaviour is partially due to a "little Putin" strategy. Little Putin strategy takes a page out of Putin's playbook ; you find areas, regions where the other side does not want to be involved completely, so you get involved 100% bcz u know the backlash wont be severe, so u cause some discomfort for the other side. Like Russia invading Georgia, the West doesnt like it, but will not engage militarily, it doesnt worth that much to them, so u look tough. Tigger obviously cannot do this at the scale RU can, that's why I call it the "little Putin strategy".

It's a way of looking tough - but in fact, it's like acting like a little bitch isn't it? Engaging in something when your security isn't truly at stake, against someone who is not a threat, and usually not in the same class. Just to get some headlines?

At least RU can argue it has some strategical reasons to be pursuing their Putin strategy, for TR, already a Western ally, the little Putin strategy makes zero sense.

U could say the dollar is strong. Since 2001 DXY (dollar vs a curr basket) mostly hovered btw 74 and 95, it is now closer to 95.

Ok.. there is some inflation: 2.0% YOY.

Next Labour leader? The psycic says Yvette Cooper.

But I think the psycic really wants a centrist ;)

#ukelection

BJ vs. Corbyn - who was the better oaf?

Exactly.

From the debate - look at those glases on Corb. He looks like the cranky high-school headmaster who has a few dead bodies hidden in the shed. It's all about rough perception. Imaging. Remember HW Bush looking at his watch during debate with Clint. Cant do that.

#ukelection

We have yet to see anyone be judged on the repercussions of Brexit. Hopefully that time arrives soon.

#ukelection

If Corbyn was decisive one way or another on Brexit he'd have a chance. They had to be decisive, and scare people - with truth (also offer some way out if Brit culture codes have such a thing). U need to hire expert politico messagers.

#ukelection

I wouldnt make a big deal abt that. If incumbent face plants, econ going bad and a Labour leader who can sell, "close" they will come back.

#ukelection

"Working-class voters didn’t trust or believe Labour. We have to change" -- The Guardian

"What the world needs is a big reset of the system where leveraged firms default, solvent firms pick up the pieces and get to earn excess returns due to their past fiscal sobriety. Since we live in a democracy, that won’t happen, instead we will have extreme fiscal stimulus in order to kick the can further down the road"

Link

The Sea Angel - wow.

"Goblin Sharks are known as 'living fossils' because they're the only living species of a lineage that has existed for 125 million years"

The Deep Sea

"@Jkylebass

China’s Companies Binged on Debt. Now They Can’t Pay the Bill...This won’t be a problem thanks to the MSCI and Bloomberg Indexes adding China. More suckers being brought into the Chinese scheme of roach motels. Investors will never get their $$ back"

Link

We live in interesting times. It gives rise to a strange new vocabulary. Like issue-non-issue. Or in finance the good news that is bad news. Watch a finance show u hear them argue all the time "this was good news, but is good news good news or bad news 😶 [for the stock market, bcz good news can cause FED to raise rates that'll lower the laughing gas, causing the market go down]".

BJ succeeded in painting others as the ones stopping Brexit, "the will of the ppl (meaning the shit cake he previously sold them on)" and rode that to victory.

Brexit.. What an issue-non-issue. US cons could only dream about something like this. It keeps on giving.

#ukelection

Link

#medicalDebt #cnbc

The fin advice part is really dumb but at least they talked abt the problem

Link

"@PGourevitch

Imagine having a passport that allows you to move freely between, live in and work in 28 countries, including many of the worlds most affluent, with excellent free health care and social services wherever you go & voting to tear that passport up and be stranded on an island"

#ukelection

I am definitely studying this subject irreverently (per Feynman) #optimization

What happens if, in a society, ppl who do not buy into the national narrative, are culturally different and focused in one region? It can happen, see Gellner who wrote the book on nationhood,

People can be categorized as tall and short, as fat and thin, dark and light, and in many other ways. Clearly, there is simply no limit to the number of ways in which people can be classified. Most of the possible classifications will be of no interest whatever. But some of them become socially and politically very important. They are those which I am tempted to call entropy-resistant.

A classification is entropy-resistant if it is based on an attribute which has a marked tendency not to become, even with the passage of time since the initial establishment of an industrial society, evenly dispersed throughout the entire society. In such an entropy-resistant case, those individuals who are characterized by the trait in question tend to be concentrated in one part or on the totality of society. The rest of this argument can now easily be anticipated: entropy-resistant traits constitute a very serious problem for industrial society.

Well, noone should run these ppl over, displace them, or destroy them. They'd need their own local governance, Scots, the Irish, or the Uighurs.

@CodeWisdom

"...the purpose of abstracting is not to be vague, but to create a new semantic level in which one can be absolutely precise." - Edsger W. Dijkstra

$749 million. Seriously.

"@AppiaPay

'...Uber paid $749 million in credit card processing fees in 2017.'. Every day #businesses are paying substantial #CardFees to financial gatekeepers and middlemen"

"In research published in Nature Communications recently, scientists from UNSW Sydney, Griffith University and Swinburne University of Technology showed that capturing hydrogen by splitting it from oxygen in water can be achieved by using low-cost metals like iron and nickel as catalysts, which speed up this chemical reaction while requiring less energy"

Link