12 Mar 2018

ಠ_ಠ!!! Freelancer


JavaScript:

  • D3

Python:

For connecting basic ML ideas to scikit-learn implementation

So this is ‘Python’: I’ve misinterpreted Python as CPython’s interpretation

Entreperneur:

  • Automatic search context by viewing behaviour? recommend background knowledge?
  • stash current job list in chrome ext (since most in chrome tabs?)
  • ‘googling’ of pdf words? useful when reading docs
  • ‘Behaviour’-based spent time estimator?

Chrome:

ML:

  • Constraint-satisfaction task
  • Constraint-satisfaction searches
  • Strong constraints (problem domains s.a. games or puzzles)
  • Weak constraints (problem domains s.a. finding the most plaussible interpretation of an image) – not ‘all-or-none’ criteria
  • Perceptual interpretation task
  • Hopfield (weight in representing the correlation of a pair of hypotheses, other things equal), for quick catch-up
  • Hopfield – for implemented library n some other insights
  • Hopfield Network – you man still need to check the original publication Neural networks and physical systems with emergent collective computationalabilities by J. J. Hopfield in 1982
  • Emergent collective computational abilities
  • Emergent algorithm – think about ‘fuzzily’ match a sentence to some meaning; think about ‘fuzzily’ match an environment to some finite-dim state
  • Delocalized content-addressable memory A general content-addressable memory can retrieve ‘What a good damn’ by sufficient partial info / words such as ‘What damn’; for more advanced ones, it may handle ‘madamn’ like ‘damn’
  • Extensive asynchronous parallel processing
  • Ising model

Vim:

Architecture:

  • Should IO/Web dominate the highest level of architecture OR should what the product does dominate?
  • BCE (boundary controller/interactor entity) model
  • And then .com bubbled (BCE was forgotten …); engineers then, sooner or later, forgot BCE
  • One of the highest goal: you should be able to test all the business rules without a web server running – more like unit tests
  • MVC (not even an architecture) was introduced in the 70s
       View
        ^  ^
       /    \
     Model (accepte View's register & update info to View) with
     BO1 (bussiness object), BO2
     ^ ^-- --------------------^
    / /  /
    Controller (accept inputs, like clicks, push corresponding cmds to Model)
    
  • Why would MVC be popular? By Bob, MVC was good; so they used it?!
  • Why could MVC be horrible? The complex nested dependencies … Check this
  • How about MVP to remove the not-welldefined boundaries? MVP?
    I KNOW WEB            WTF IS WEB? EDIBLE?
                    ||
    Presenter -------> Boundary <- Interactor -> Entities
      |     \       ||                |
      v      \      ||                v
    View data \-----------------> Response
      ^             ||
      |             ||
    View (module)
    
  • Pros: we can test Presenter, Interactor, Entities without Web; componets on the LHS always have only inward arrows to the RHS – they behave as plugins!
  • Bob: Plug it in!
  • Source code control systems, such as Git, actually stores transactions (commits) rather than states
  • Database is a detail!
  • For any details, we’d like to treat ‘em like plugins!
  • Bob: A good architecture is one that allows you to defer decisions (decisions about UI/DB/Frameworks; an awesome architect maximize the number of decisions that could be made later

Tests:

  • Challenge: automatic unit/integration/acceptance tests?
  • Integration tests: ‘across-entities’ tests, e.g. some function + test DB
  • Acceptance tests: app/product-level tests; perhaps Selenium for web app?
  • A decent comment, and I quote here, “If you feel it’s not easy to write a unit test, chances are it’s not a unit test at all.”
  • TDD: test before implemtation
  • BDD: behaviour over implemnetation

Murmur:

Songs: