Rainy_One

 

Main

 Found the use of this mysterious bit of python code:

It is instructions to the interpreter to run this file as the main one. Because a project

can easily have more than one .py file in it, and one needs to check if it is one that is

imported or one doing the importing...

                                                                *     *     *

The underscore, in Python, is generally used to print the last output. It can

also bee used as a nameless variable.

By convention, coders use it to tell later users of some code to keep something private.

Pythonesque

 Been trying migtily to get a grip on python: multipurpose language,

object-oriented, blah, blah. Then I saw a quote from G van Rossum wherein

he said he had developed python to faciitate working with data. That

makes sense. 

C is arguably the fastest language to get things done, but it is long-winded and tedious

to write. python is breezy but its interpreter is written is C, so it will take comparatively

longer to execute. The interpreter translates python code to bytecode, which then

runs on a virtual machine. Re bytecode:

'Byte code is an intermediate code between the source code and machine code. It is a low-level code that is the result of the compilation of a source code which is written in a high-level language. It is processed by a virtual machine like Java Virtual Machine (JVM).

Byte code is a non-runnable code after it is translated by an interpreter into machine code then it is understandable by the machine. It is compiled to run on JVM, any system having JVM can run it irrespective of their operating system. That’s why Java is platform-independent. Byte code is referred to as a Portable code.'

Geeks for Geeks

As for what lives where, Python3.9 comes with libraries. Modules are functions, groups of

related modules are packages, and packages travel as librairies. A package has an __init__

functionthat calls the appropriate module. Pip install will make move a module's contents

to an easily accessible file.

As an exercise, I am importing the sys module and asking for a print of sys.path. i will see

where the machine will look for  sys functionality as needed:

Stingy

 Corpus0 is he simplest case, with diagram showing links below.

How should it pagerank.

I re-did the two approaches, sampling and iteration, but with much

smaller number of cases. There is even more difference...

With  10 000 sample for approach 1, and .00005 threshold on 2.

Going stingy with 100 sample, and .005 threshold on 2.

P_Rank

 Anyone who uses Google has some notion of pagerank: it is the algorithm

that for any specific request, ranks the most popular sites. The first assignment

of the uncertainty lecture aims to code for a pagerank.

It may be easy enough to follow users from site to site, but describing such

a system gets bogged down in sollipsistic cycles. There are some sites that only

lead to a few others,- like apple pie recipes - so that randomizing algorithm risks running

in circles. To counter this, a dampening factore of 0.85 probability places the program 

in popular sites, and 1 - 0.85 probability covers access to any page, including

the one we are on.;it is always possible to get out.

One writes the transitions code, a sampling approach (10000) and an iteration.

There are three corpus sets to test the code against.

I have been using the code from:

https://github.com/lashewi/CS50-AI/blob/master/pagerank/pagerank.py

Will be working through the code tomorrow...

Markov

 Starting off this post with a photo of Russian mathematician Andrey Markov(1856- 1922)

for the very simple reason that he was rather good-looking.

His area of interest was stochastic processes ( ones where randomness is involved); and he

is known for Markov chains, essentially the idea that, as the Russian-language Wikipedia entry put it -

the past does not influence the future. Below, a Markov chain with probabilities attached for

each one-step alternative:

                                                           *     *     *

Nothing terribly surprising in this model. Starting on a 50/50 chance of rainor sun, each

run prints out a plausible sequence...

                                                            *     *     *

The second model introduces a bit of a twist: it is the probability of sun followed by rain

that is given! A camera captures whether emploees are entering the building with an umbrella.

I ran number 2 a number of times, and the results are always the same.

Flowering

 Took one of my plants in a few days ago - a desert plant -, and it has flowered.

It doesn't flower every year; thought I'd share...

 

Bake It

 Pomegranate isreally wonderful. One can build up a model

very simply. Below, the train model:

Once all ia done; bake()

One is then free to run different types of queries. The first, inference, wants to

know if we are going to make the appointment, given that the train is delayed.

Computer gives back the odds on that, pure and simple. Note that the only parent

node to making the appointment is the train.

The likelyhood code below makes a different kind of enquiry. It wants to know if I will

make it on a day where the train is on time (0.9 prob) but whether I will exhibit this

virtuous behaviour on a no rain day with no maintenance. The probability of all that is

a paltry 0.3402 !!

A third type of query possibility is sampling. The computer makes multiple runs on 

conditions.

We are interested about making the appointmnt given a delayed train (proba 0.6)

but we are catching that the train maybe delayed as a function of rain and maintenance 

(and indeed mainteance itself is inversely affected by rain). 

                                                              *     *     *

Alpha

 Baysian analysis allows for joint probability, below:

If I add the two columns, I get back to the 'rain in the afternoon' set of probabiities.

If I add across rows, I return to the 'clouds in the morning' numbers. So what is likely to

happen on any given day: no cloud and no rain; cloud and no rain; cloud and rain; 

no cloud and rain. In that order...

                                                             *     *     *

I can also ask a 'time-inverted' question. What is the probability of having experienced

cloud in the morning , given it is raining in the afternoon.

Alpha, here,  is used to produce a standard probability distribution that adds to 1!

                                                           *     *     *

Just nstalled pomegranate, a probabilistic programming library for python.

Tomorrow...

So Done

 So, what kind of magic is this!? What Bayes' Theorem is addressing is

conditional probability. Math is Fun gives us a clear view of what this concept

refers to: the probability of an occurence, given a previous one...

The two draw events are independent, but the conditions have changed:

                                                                     *     *     *

How Bayes' T can be derived:

*     *     *

Interestingly, Thomas Bayes(1702 - 1761) was a compatriot of David Hume(1711 - 1776) who

- famously - defined causality in experiential terms: the constant conjunction of events.

                                                             *     *     *

Wiki Help

 Biting the bullet on Bayes' theorem. This seems harmless enough,

but it  is the basis on which links are made between medical test

results and disease...

Thank Heavens for Russian Wikipedia, which presents the argument as follows:

Dystopian

 https://www.theverge.com/2021/10/19/22735612/facebook-change-company-name-metaverse

https://decentraland.org/

Painful

 Looking to start work on the Uncertainty lecture from

CS50 AI. 

It will involve Bayesian analysis:

My brain hurts already...

Computer Play

 How the computer plays: it plays quickly!!

                                               *     *    *

Benn playing with changing the number of mines in the game (line 9 on runner.py).

Below, a game with 10 mines led to this punishing outcome on the last move.

Ai had been on a run of safe moves up to there...

So there is an element of chance that can sneak back in, and hence one won't always win...

Another mines10 that flips into random mode, even with a move 1 to a 0 tile!!

Loss on eight:

Works!!

 Finally found a version of minesweeper code that works

(on PyCharm). Will be examining code tomorrow:

https://github.com/MoosaSaadat/minesweeper

Digging In

  Comes down to learning python and AI at the same time.

For python functions:

https://realpython.com/python-return-statement/

Been following along with this youtube :

https://youtu.be/EhLYArPYIUk

Makes perfect sende: every cell that is a mine is worth 1,

every cell that is safe is worth 0.

TDrop

 Getting ready for the big temperature drop this week-end.

Back to seasonal,essentially from daytime highs of 20 to  teens.

                           

The Challenge

 I had forgotten just how challenging this game is. Below, some modest

beginner solutiions...

https://cardgames.io/minesweeper/

Srarting to see where the computer will outclass me: solved this last one

at the very end by considering how many mines were left to be played.

Such a procedure would be mindbending on a larger board...