Discussion is telling any idea.

• moreover, in academic context, discussion also refer to the activity of telling any idea in written form and pass it to peers, so that the idea goes through the peers(for review and sharing).
• sometime it also refer to the activity of multiple people telling their ideas on the same topic.
• foundation knowledge

optimization problem where in the constraints, all occurances of input is linear(to the power of 1).

• example: \[$ f(n) = n \\ constraints: \\ n < 5 \]

The noblest kind of beauty is that which does not transport us suddenly, which does not make stormy and intoxicating impressions, but that which slowly filters into our minds, which we take away with us almost unnotioced, and which we encounter again in our dreams; but which, however, after long lain modestly on our hearts, take entire posession of us, fill our eyes with tears and hearts with longing. What do we long for at the sight of beauty? we long to be beautiful. We fancy it must bring much happiness with it. But that is a mistake. – Friedrich Nietzsche, “Huamn, All too Human”, 149.

85% law is an heuristic of “difficulty sweet spot”.

85% law states that highest learning rate occurs when:

• success rate is about 85%
• \(\frac{\text{success rate}_{\text{before practice}}}{\text{success rate}_{\text{after practice}}}\)

By adjusting these difficulty parameters, you can affect your success rate. Make the task harder when you succeed too much, and keep the success rate around about 85%, that is about 5 success in 6 tries.

difficulty parameters

There’s handles you can adjust to affect difficulty in every practice. It’s in the form of adjusting the level of support you can consult.

define success

make a quantitative measure of success, and thus success rate. there are multiple ways for 1 activity.

activity	difficulty parameters	success definition
solve prolbem	open book?	5 right answer in every 6?
	worked example?	time?
	cheatsheet?	effective step?
	without anything?
drawing figure	reference?	5 looking good out of 6?
	cheatsheet?	time?
	tools?(spray, distortion)	effective step?
	values?	poses?
	without anything?
programming	without mouse?	5 test success out of 6?
	arrow key?	5 perfect function in 6?
	cheatsheet?	5 perfect use case in 6?
	examples?
	lsp?
	design doc?
	without anything?

(no term)

using 85% law on programming (where detailed scheme of difficulty parameters and succes definitions are elaborated.)

• source: https://www.scotthyoung.com/blog/2020/04/05/circle-or-stairs/

The main difficulty parameter is the context. Namely, the environment, how much complete the puzzle have already become.

• musical example: It’s harder to remember the lyrics, lines and staging and simulate it in your mind, it’s easier to be there (in the show)and say the lyrics/staging.

That property of 1 linear optimization problem being actually 2 linear optimization problem.

• navigation note
• compilation note
• concept building with Zettelkasten
• data-interpretation-synthesis layer model with Zettelkasten

Most system with autonomous control have self-adjust functions to maintain a given state.

This kind of control could be calculated by enginner and implemented as design, or have some learning function that adjust the adjust parameters itself.(like acceleration on some input).

To maintain a ball in a zone, an acceleration is applied based on coordination and speed.
This acceleration could be calculated by definition of zone, and some equations. Or it can be calculated with machine learning algorithm.

When full environment information is available, a deterministic normal algorithm would suffice.

When partial environment information is availabel, or information too complex, a learning algorithm may be better.(but not necessarily)

It is worth noting that learning on time sequence is emphasized in this kind of learniong algorithm, as input-output for the target function(cause-effect relationship) is given in time sequence of sensor inputs.

For tasks where the target function is clear and universal, machine learning is redundant. You don’t need a machine learning algorithm to know that \(F = ma\)

For tasks where the target function is hard to tackle because:

• target function is dependent on environment(personalized) song recommendation, personalized assistance prompt(time,location and app), session specific query(chatgpt session)
• target function is too complex (thousands of parameters/influences) language-meaning

machine learning could help guessing the target function autonomously(design the learning algorithm once and use everywhere.)

observation 1

I used to like jetBrains’ logo and style, but now I hate it

observation 2

I used to like my-emacs-config’s theme, now I kind of hate it

observation 3

I used to like machine(as with Ray Daleo), now I kind of hate it because I ruined it for myself

observation 4

I used to like Doom emacs’s startup, after a while I kind of hate it because there’s a while I see it and not doing meaningful stuff, and now I like it because I’m back doing meaningful stuff with it.

observation 5

I have absolutely no opinion on the minimal startup emacs(which is empty scratch buffer), or vim(which is either straight to file or a line of text signifying it is vim.)

hypothesis

minimallism reduce association of idea-visual with feelings. That could be 1 upside of minimallism: distraction is reduced and you see clearer the real stuff.

If you only create new org-roam node via link literal operation(make a [[roam:title]] link and click/follow it.), then the graph is guaranteed to be connected.

It could be proved by induction: base case being only 1 node, and doing this make it connected, and recursive step being adding a node and a link to a linked graph, which make them a linked whole.

Also note that if you only create new node with links(that is, every link is leading to a new node), there won’t be cycle, and thus would be hierarchy.

RSA is an asymmetric encryption method.

keys

you are given 2 keys in RSA, a public key and a private key

public key

public key consist of 2 numbers, \(e\) and \(n\).

private key

private key consist of 2 number, \(d,n\).

Encryption

Encryption is done with the public key \(e,n\).

1. You translate the message into an really large integer \(m\).(e.g. concat all ASCII code of the characters)
2. \(cipher = m^e \mod n\)

Decryption

Decrypition is done with private key \(d\).

1. \(m = c^d \mod n\)

Key generation

\(e,n,d\) are generated by

1. take 2 very large prime numbers \(p,q\), let \(n = p \dot q\)
2. select some \(e, \text{where } 1
3. \(\phi{n}\) is count(number less than n that coprime n).
4. \(\phi{n} = n-1\) if n is prime number
5. \(\phi(n) = \phi(p) \cdot \phi(q)\) if \(n = p \cdot q\)
6. calculate \(d\) such that \(d \cdot e \equiv 1 \pmod {\phi(n)}\)
   • as \(e\) prime \(\phi(n)\), \(gcd(e,\phi(n)) = 1\)
   • which is equivalent to \(e \cdot x + \phi(n) \cdot y = 1\) \(x,y\) could be computed with Extended Euclidean Algorithm
   • which is equivalent to \(e \cdot x = (-y) \phi(n) + 1\)
   • which is equivalent to \(e \cdot x \equiv 1 \pmod {\phi(n)}\)

• resourse
  • this wiki is clear and informative
• reminder
  • There’s a lot fuzzy joint needs rigid proof in this note. Establish them.

• one fun fact about human mind is that it is totally not rational.
• you see a stuff, and immediately a framework, an idea, a structure appears out of blue.
• a dangerous part is the implementation plan. Implementation usually take time and focus, so if you work on your implementation plan without recording them on an external media, it’s going to be forgotten or get distoreted.

Zettelkasten is great way to settel rustleing ideas in mind and build sound structure in them. It is likely that you want to do anything with zettelkasten(at least at the start, when you are collecting ideas and building design)

for different projects, it’ll gradually become imperative that you don’t want them to be mixed together, majorly because the truth stand of each may be different:

• for revising a module, you care about the specific system that would be used as correct answer. You don’t necessarily believe in them, or find them useful/interesting
• for building a software, you care about your own abstraction and designs that are present in the software along with tech references of how you can implement them They are not true stuffs, and tend to be large in quantity
• for noting down your thoughts, you don’t really know whether it is very stupid or worthy a second look. The soundness of them is in doubt.

The name-spacing require 2 operations:

• search with a name-space
• search without a name-space

There are some ways to do name-spacing with zettelkasten.

One way of perceiving usage of tags and links is:

• tags are gates to a building
• links are corridor navigating around once you are in

e.g: In Zettelkasten, you want few tags, few notes with tags and lots of links between notes. You can tag some navigational notes but not the notes they link to. In browsing, you want to read your notes about “machine learning”, you could search for the tag, and notes that are entrance to clusters of you machine learning notes would appear, and you could explore each cluster as you want reference: this comment

reading an article is a procedure, and naturally a todo task if planned.

Reading seq model defines a Sequence data structure with 2 methods.

Sequence

is a seq with elements all object/links to reading materials

add method

how to add to the seq

consume method

how to consume element(read stuff) from the reading seq.

if one such entity is implemented, should reduce the amount of operation required to “add to reading list” and “read an article” to be one

Minimal operations is concept for algorithm or user experience, that there’s a minimal operations to achive a goal.

example:

• in Euclidean Algorithm, you have to go through all the bits of the 2 numbers to find their GCD. There’s no way to find the GCD without going through them.
• with text editing, the minimal operation of each text operation, like “copy this line from 2nd to 9th letter”, is 1 key press(define a function that do this stuff and bind it to a key). however, if you have too many procedures in this manner, you’ll most certainly run out of keys, at least the easy-reaching ones, and soon require paging/navigating, which would be implemented to another key, or some prefix system. And that would be y a w like vim.

integration:

• 21
• 22

mental load refers to the mental effort made to memorize or recall the command for operation. For instance, to remember a shortcut and what it do in a software.

Mental load could be measuered with number of spanning rules. In the worse case, the number of spanning rules would be the same as number of operations. The coresponding scenario is that there’s no structure and the association of keys and procedures are random, in which minimal operation could be achived

Mental load could be reduced with prefix/hierarchy system, visual aid, or verbal system, which trys to represent bulk of operations with few spanning rules.

For instance, in vim key-binding, verbal system is a major component. with y [motion command], a large set of copying commands is covered with mantal_laod(motion commands) + 1, like copy 2 lines below, copy 2 lines above, copy until next “f”, copy to end of line, copy till next word…

downsides: note that visual aid could be distracting, verbal system could be hard to implement, and hierarchy system is sparse and could be not too performant.

the largest number of minimal operations in a command system is equivallent to the size of longest n-hoffman code string that could represent the all operations in the command system. n refers to number of keys. n-hoffman code refers to hoffman code with each digit from set of all available keys in the command system

It is easy to see that n-hoffman code is the smallest string that allows deterministic parsing to all the operations.

with n-ary (digit system with n options each digit) representation:

• n-ary representation make every string the largest number of minimal operations
• with n-hoffman code prefix tree, could assign shorter codes frequent procedures, count all minimal operations and other that reflect real-life use.

self-evident/instinctive is for a phenonmenon that, for a process, you could recall the whole process with ease immediately, on seeing any part of it.

• like packaging a box: you see the tape, you know how you would wrap it around the box.

Note tha self-evdience is related to training. Some process are self-evident to some person, some don’t. For man who see tape for the first time, it may not be so self-evident how to use it. I don’t konw how to operate firearms, but for military man the guns and firing would be self-evident.

I have a course. I have a path. There’s stuff that I feel right doing. That is what I refer to as “myself”. Stuff I think and feel right, stuff I do and feel right. Game a lot don’t feel right, think low of myself don’t feel right.

And if I don’t feel right about the 2, the mental or the physical, I lose confidence, I lose integrity, I lose sincerity. They raise doubt, they raise stress.

And How do I lose them? I don’t give them up myself, at least not consciously: There’s distractions, and I followed them. They may be primitive, that distracts me out of the nature of the very human I am and the nature of the distraction: I’ll always prioritize eating when I’m hungary; or traditional, that I have allowed them in my life in the past and they thus stayed.

And the traditional distractions is they main enemy from being myself and being confident.

this is a criteria of How small should my GTD action be. Generally, in GTD we want full capture, full establishment of action, and no partial note. And we want the action to be easy to shift focus to, meaning no extra planning, no extra learning/research of how to do it: they should be done seperately or in determining the task.

Therefore, if the desired outcome is self-evident, meaning that you can immediately see how you make the outcome, you achived the 2 criteria the same time:

the information is complete

you know exatcly what to do from beginning to end.

easy focus shift

you can load the entire context (or most of it) immediately from the description.

observations/data

1

One time, I read paper and I could keep track of threads, stuffs I noticed, can skim through passages and tag them what they are talking about

2

One time, I read paper and even finishing sentence is hard. I can’t really keep track of things in my head

parameters

Hot water

drinking it.

Hot water spa

a thing I do. I blow hot water for a while and let it steam my face. get me to calm down.

Music

type of music

Happy? Musical?

Noise

Environment

stance

standing, sitting

exercise ahead

psychological factor

deadline?

There should be a cheatsheet(better physical) available for each procedure that I’d perform, like literature review, reading paper…

• establishing procedure should produce cheatsheet other than the main document [2023-05-13 Sat 16:30]

It apppears that it is more lovely to see:

• people actually confident doubting themselves

than:

• people acutally inconfident, insecure trying to act confidently [2023-05-21 Sun 15:20]

confidence is a self-preserving state:

• by mimicking being confident, you’ll approach actually confident
• by being confident in part of something, you’ll approach confident in the whole of that thing
• by posing a subset of confidence signals, the other would appear automatically

[2023-05-21 Sun 15:22]

data

me using RH timer for pomodoro instead of org-pomodoro or org-clock-in stuff

interpretation

distributed mech

pomodoro is a mechnism to maintain level of focus/energy in a period of time. There’s no need to keep global control of it as long as it does it’s work and make the focus of that time period sharp.

time-distributed

Treating various time-periods in a day like asynchrnous nodes for processing

un-clockable activity

when doing knowledge work, sometime it’s hard to define what to do, as “finding out what to do” is the work, or the task would be “process this batch of information and make some structure out of it”, which may not success(as there’s no such structure). A large timesink activity like “process information on comp212” may be practical.

why RH timer

It have neat GUI, can stick on top, have pomodoro shortcut.

[2023-05-21 Sun 16:52]

data

I want to be able to play something in jazz improv/comping. I want it so badly I collect everything and practice them.

interpreation

I may end up only be able to make common gibberish, went on autopilot and only play stuff I heard of. I saw that before with my 6-5-4-3 improv.

more

being able to talk all the time is a skill very much wanted. but don’t I want to speak only the gems and not discarding junks?

[2023-05-22 Mon 11:37]

it’s important you have an expectation of what you will be making, so that you know where to build off and notice special ques and signals, stuff out of ordinary. can do it by asking for other’s practice tracks and do some little study [2023-05-22 Mon 11:55]

• for exam -> remembering a marked one right answer

• for research/skill -> make stuff that make sense to you

  [2023-05-22 Mon 12:15]

desired outcome

a function that can use and update file-variable in an org-capture session

scenario

in licks.org, there’s an count-id(Lick No.3). How I could implement that with org-capture automatically fill in the headline [2023-05-22 Mon 12:24]

improvisation on a number could be seen as research on a topic and have your own argument. You could just do it directly, or see some literatures of how other people think of it, and maybe build off that. e.g.: it is surprisingly hard to improvise on autumn leaves, and most recordings on spotify didn’t do much better than I do. But skimming through, you’ll find some valid ideas [2023-05-22 Mon 12:31]

the syllubus would consist of:

a Herman No.2 Piano skill set

composition of some fundamental playing skill

a Herman No.4 music research/discussion skill set

composition of some research skill for the student discuss numbers, their thoughts, and develop their arguments/own insight and implementation of musical ideas [2023-05-22 Mon 13:47]

desired outcome

write a tutorial about how macOS work and various APIs it exposes that one can use with program

https://developer.apple.com/library/archive/documentation/Darwin/Conceptual/KernelProgramming/Architecture/Architecture.html and maybe write some C to test it, and relate them to my usage I may stick with my macbook for quite a long time. So I hope at least I understand what I’m using.(it’s not like I’m ever going to get a linux machine) [2023-05-31 Wed 02:51]

a paradigm is that, instead of a master theory, a master paradigm, which should incorporate different techniques and be consistent of various observations, use very scope-specific cluster of ideas.

For example, one can approach music as there’s one correct theory of composing/instrumentation(master theory) and translate their techniques and thoughts into arguments compatible with that theory. The localisation version may be approach music as various techniques, observations and ideas, each with own data and interpretation, maybe integration. [2023-05-29 Mon 17:16]

it is important that you make creative/opinion decision on the thing, and finish your creative project(that is a project that is to span/implement/illustrate/expriment your opinion/idea). That way you remember of a simluation of how things may went [2023-05-28 Sun 01:29]

with a general simple rule, scripts/program is likely to meet special scenarios that work unexpectedly Strategies to mask these failures are:

• specify the failure scenarios, and advise the user to avoid them
• use guard to avoid the scenarios in program (require strategy to deal with them) it’s similar to that in behaviour tree, you can add if-then structure as first child of an or node, to mask latter children’s fialure(if they are too complex to be modified)
• re-design with the failure scenario considered [2023-05-26 Fri 12:35]

Since bookmark are for later refrence, and is a type of capturing, it totally makes sense, and make me happy(don’t have to remember where and how I do bookmarking all the time) [2023-05-26 Fri 12:06]

literary programming and automation are alternatives sometimes. Some time you need a script to automatically find a class of object(like pdf files), because it’s too much trouble keeping them mannually in a text file. With literary programming such as org-babel-tangle, such burdon is elevated (when scale is small), as you can have very easy-to-understand code snippets and easy/intuitive way to run and navigate around them.

They are not entirely the same, but sometimes interchangable [2023-05-26 Fri 11:47]

• write down possible options.
• someties you’d find simple, stupid, narrow, domain-specific ideas way better(to write, to use) than generalized, complex ones. [2023-05-26 Fri 11:44]

[2023-05-23 Tue 22:44]

western? some parts. Some Guy Rich similar atmosphere. Snare and the horn how can I make something similar… [2023-05-24 Wed 23:56]

wistle accompanying snare, or very dry acoustic guitar [2023-05-25 Thu 00:17]

driving focus off. https://music.youtube.com/watch?v=XaAiNL-YJIs&list=RDATfdYmVib3Az [2023-05-22 Mon 14:45]

data

there’s no known way of recovering a losed arm. There are people working on it, but the bulk of us mainly try to avoid losing an arm till the day we die. [2023-05-22 Mon 15:28]

These well-renaissant notes. so good [2023-05-23 Tue 14:05]

Itis obvious they have their own strengths/weakenss:

• pen&paper
  • slower to write words
  • can draw
  • anotate really easily
  • really easy mindmapping
  • good for drafting strucutres
• org-mode(aka text-based computer note-taking)
  • faster(words)
  • can be searched
  • manipulation
  • completion
  • links
  • draw is really hard
  • anotate slower,
  • structural modificaiton/implication is really hard(only hiearachycal is intuitive)

[2023-05-25 Thu 13:52]

scenario

what should we do if we encountered bullying

(no term)

ask - is justice the most important thing to you, is payback, or the balance that important to you that you should focus on that entirely

• if not - always do the most important thing first
• if yes - go fight as violent as you want

[2023-05-25 Thu 19:27] %

to take up the task from last time, either:

• have a global picture where the previous work is visible and it is self-evident how they should be processed(the vision is clear)
• GTD task traking
• for some knowledge work, it is hard to have precise task definition. Therefore, a mangeble-sized workspace inside a more general GTD task may fit usage
• project zettelkasten
• navigation/index note

IT’s an practical book. I looked the menu, and saw some insights that looks really delicious [2023-06-04 Sun 18:30] Sweller et al. - 2019 - Cognitive Architecture and Instructional Design 2.pdf: Page 1

[2023-06-05 Mon 16:52]

I want to be working on an abstraction level higher than ANN so I could incorporate ANN, SNN and other stuff, and be doing the definition/idea rather than technique [2023-06-07 Wed 10:40]

[2023-06-14 Wed 16:16]

def of confidence

a lot, but they try to describe a common feeling that occurs when man’s flesh and soul are in the same place

• feeling/consciousness of one’s power
• feeling/consciousness of reliance on one’s circumstances
• belief that one will act in a right/proper/effective way

The effect and existence of confidence is self-contained

exotic sound. [2023-06-26 Mon 00:01]

on C, Dm7 give a dorian sound. The Dorian sound on D converge to the C major scale, the C majorness [2023-06-26 Mon 00:14]

• don’t write it down: it is how they played it that matters, it is how you heard them that you recall in a session.
• use your ear. audio clip to copy from is best. [2023-06-26 Mon 15:45]

• I’ll have a book of techniques. a technique library
• when I want to abtain a technique, I add it to my practice diet, with specific “what to practice(content of practice)” and “outcome”(where I remove it from the practice diet)
• in each practice session(I’ll set default at 30minutes), focus on getting a small portion of the technique right and instinctive, make myself familiar with it and use them flexibly. revisit it very soon, cycle on small subsets of it
• from time to time, assess the practice diet, and discard already familiar technique and add new technique I’v seen and is interested at but don’t have time for it before

[2023-06-26 Mon 00:15]

in the following heading, when Samsonovich :: xxxxxx is followed, would open a buffer *select-link* in help mode. Removing :: from the description would remove the behavior

That is work from crashed by empty sky [2023-06-29 Thu 11:51]

discipline is a way of organising ideas and theories, as well as research object like neuro circuit.

Like other organisations, disciplines would overlap, and can be quite arbitary

[2023-07-29 Sat 22:37]

• policy is a map from state to action (as defined in Markov decision policy)
• neural network can be abstracted to a map from a vector to a vector
  • input vector as the state
  • define the choice of action on the output
• –> neural network map state to action
• –> neural network as a representation of the policy
  • other representation model: desicion tree…