Friday, April 9, 2021

Dash cam landscapes

There was a time when I watched Russian dash cam videos.
Many of them ending in horrible accidents.
The tumbling car would come to rest and I would see a small place in great detail.
In silence.

That I - in the Netherlands - would see that one tree, that specific blade of grass, that lonely roadside weed. They were recorded, they will not be forgotten.

I feel the earth move - I feel the tumbling down - the tumbling down.
(Philip Glass - Einstein on the beach)

I searched the internet at random for other fitting comments. 
I have no talent for good texts, but I can search for them. This is what I found:

We drove approximately 400 km by automobile from Chelyabinsk through Stepnoie to the Arkaim and recorded vegetation structure at random by visual observation along major traffic roads running with automobile approximately 10 km in distance.
The types of vegetation depended mainly on physiognomy with relatively smaller scale approximately some hundreds square meters: Pinus forest, Betula forest, cultivated land and pasture.
Plots along the major roads were randomly selected to cover representative vegetation pattern. Further, around the Arkaim, major vegetation types ware described. Field observations were carried out during early summer season, middle June in 2008.
This theory is contradicted by the 12 hours I spent yesterday watching Russian car dash-cams on YouTube. The russian car inventory must turnover at least every 6 months. Literally and figuratively. (3 stars or bust)
There’s a deeper point here: if your view of Russian life is conditioned by what is available on YouTube, you’e asking for a pretty skewed view of the country. The people who post dash-cam videos are, by definition, those who have cars, dash-cams, and internet connections. Faberge eggs are very pretty, but they say relatively little about per capita wealth in Russia on the eve of the revolution. (Ckb)
As we approached the city, we saw:
  • a billboard advertising a concert,
  • large signs on top of the buildings near the railway,
  • the cathedral on the hill,
  • the towers rising from the horizon gleaming in the sun,
  • the skyscrapers thrusting, soaring, sitting above the rest,
  • a crowd in front of the showroom near the highway,
  • more and more people along the way,
  • very ominous dark clouds in the sky,
  • orchards and orchards,
  • the high and snow capped peaks.
She proposes that the brains of all six species have gotten bigger because humans have radically changed the landscape. Where there were once pristine forests and prairies, there are now cities and farms. In this disrupted environment, animals better at learning new things were more likely to survive and have offspring.

Sources:
The video:
https://www.youtube.com/watch?v=n-VymQeFoVQ
Dash cams:
http://animalnewyork.com/2012/russian-dashcam/
http://coolmaterial.com/roundup/10-craziest-russian-dash-cam-videos/
http://www.wired.com/2013/02/russian-dash-cams/
http://www.dailydot.com/lifestyle/russia-dash-cam-videos-livejournal/
Landscape flora and fauna:
Vegetation landscape around the Arkaim eco-preserve, southeastern Ural, Russia
NYtimes - as-humans-change-landscape-brains-of-some-animals-change-too
Dash-cam comments:
http://marginalrevolution.com/marginalrevolution/2015/02/sentences-about-russia.html
Found poetry:
Bing search for "as we approached the city"

Sunday, April 4, 2021

Beta distribution - learning statistics with Python

Why? 

So I'm reading this nice book on Bayesian statistics:


I'm in the chapter on the Beta distribution, and things start to get new and difficult for me:


So I follow the advice of Nassim Taleb and start playing around with some Python code. I'm not a rich guy like Taleb, so I don't have Mathematica. And this is too complex to do for free in Wolfram Alpha.

I copy and paste a crappy Python program and things become wonderfully intuitive.

Sampling and the beta function

I make an urn with 1000 good balls and 9000 bad balls.
I draw 5 samples from the urn.
I make a histogram and the corresponding probability distribution.
Then I plot the beta function for each draw. This shows the distributions that were most likely to produce this draw.
With increasing sample size both the histograms and the beta functions converge to 0.1. (Duh!)

Sample size 10:
Sample size 100:
Sample size 400:

Is 400 enough?

Many years ago I read this excellent book:


It says that a sample of 400 is sufficient in most situations. For the hypergeometric distribution that we draw from this seems to be the case. Going to 1000 samples does not seem to add that much certainty anymore.
The crappy Python code

import numpy
import matplotlib.pyplot
from matplotlib.pyplot import hist
import seaborn
from scipy.stats import beta

ngood, nbad, sample, runs = 1000, 9000, 1000, 5
samples = numpy.random.hypergeometric(ngood, nbad, sample, runs)
print(samples)

seaborn.distplot(samples)
matplotlib.pyplot.show()

# Generate the value between 
x = numpy.linspace(0,1, num=400)

# Plot the beta distribution

matplotlib.pyplot.figure(figsize=(7,7))
matplotlib.pyplot.xlim(0, 1)
for s in samples:
    matplotlib.pyplot.plot(x, beta.pdf(x, s, sample-s), 'r-')

matplotlib.pyplot.title("Beta: " + str(sample) +
                        " samples from " + str(ngood) +
                        " good " + str(nbad) + " bad" +
                        "\n" + str(samples))
matplotlib.pyplot.xlabel('Values of Beta Dist.')
matplotlib.pyplot.ylabel('Probability')
matplotlib.pyplot.show()

Friday, April 2, 2021

From a strange planet - 22

This time we don't take a road trip along many webcams. Instead we take a detailed look at  Fv296 Filefjell v/Varden and its surroundings. We use many open sources but mainly Google Street View.
We see three objects: 1) an electricity pylon, 2) a container and 3) a sign with a ladder. It's a mountainous terrain with sparse bushes. The rest is hidden in white mist. It looks empty and far away from everything.
But it's not that deserted if we look at a topographical map. There are many clusters of houses, inhabited and accessible by road.
On the satellite picture it doesn't look neglected. But it does look inhospitable to vegetation. Bushes, grass, nothing has an easy foothold here.
Winter is always present, 
even in summer the landscape looks sparse and transparent.
Snow remains in the background.
And very deserted and mountainous. Google Streetview
This is the kind of landscape I dream about.
A small hut with a porch, from where I can see the mountain every day.
Far away and in solitude.
But there is no real solitude here.
The garbage container is a sign of life.
Local people make pilgrimages to it.
They don't make pilgrimages to the sign.
I've never seen someone climb that ladder.
I've never seen anyone just looking at the sign.
But it's beautiful, like a modern Stonehenge sigil.
People do walk here.
And they walk their dogs.
But they don't notice the sign.
At night mysterious lights appear.
Sleeping houses come awake.
I sit on the porch of my simple hut,
and wait for the lights to come on.
For just 134.231 Euro I could buy an appartment in the neighborhood.
 At Nystøgovegen 500 FILEFJELL/TYINKRYSSET
It would be far from the shopping centre.

Sunday, March 14, 2021

From a strange planet - 31

A pleasant sunny, snowy road.
But what is that square, dark shape in the background?
What are those rectangular boxes?
Spooky geometries, both by night and by day.

Menacing cars parked in the wood.
A secret transaction?
Did someone walk into these woods? Did they come back?
A lake is watching through the trees. Who drowned there?
Will this place be pleasant when spring comes?

A small hill.
Probably it's very old. Older than the road.
It might hide a buried treasure.
Or it could be solid granite.
That's for you to decide.
Why dig here, if you can dig everywhere?

The horizon calls. Very horizontal.
The house shines.
Rows of trees. Very vertical.
Rotation preserves angles and lengths.

Fields wait for spring in the background.
Looking at fields is a blessing.
Time slows down. Calm descends.
Thank the creator for fields. They save us daily.

The obvious road.
And the less obvious road.
Going there, and there.
None go to the dark mountain.
The radio tower transmits, but not for us.
Someone owns it, but not for us.
We drive by, obedient.

Nothing happens, nothing changes.
A parallel road no one ever takes.
A fence separates two boring spaces.
Black mothers with black strollers.
Roll new life across danger.

It is not how it looks.
Thomas Ligotti is here.

Far away is beautiful. Very far.
The woods are complicated. Light scatters.
Who or what lives there? A statue?


Borderlands.
Trees like flags. Clouds like flares.
Shining roads.
Conveyor belts to home.

Hills.
The horizon of longing.
Thirty years war.

Fv. 289 Hurum
https://www.vegvesen.no/public/webkamera/kamera?id=1330136

Fv. 281 Storsand Hurum
https://www.vegvesen.no/public/webkamera/kamera?id=2046858

E6 Nøstvett
https://www.vegvesen.no/public/webkamera/kamera?id=2580261

E6 Støkken
https://www.vegvesen.no/public/webkamera/kamera?id=2037119

E6 Vestby
https://www.vegvesen.no/public/webkamera/kamera?id=278609

E6 Ingedal
https://www.vegvesen.no/public/webkamera/kamera?id=109904

Fv. 118 Svinesund
https://www.vegvesen.no/public/webkamera/kamera?id=457918

Fv. 101 Prestebakke
https://www.vegvesen.no/public/webkamera/kamera?id=110917

Fv. 124 Strømsfoss
https://www.vegvesen.no/public/webkamera/kamera?id=109117

E18 Ørje grense
https://www.vegvesen.no/public/webkamera/kamera?id=722047

E18 Holstad
https://www.vegvesen.no/public/webkamera/kamera?id=627532