Skip to content

Latest commit

 

History

History
100 lines (81 loc) · 3.42 KB

DEVNOTES.md

File metadata and controls

100 lines (81 loc) · 3.42 KB

Early Findings

  • finer auto-correlations mean easier connectivity inference
  • how you normalize matters more when signals are more different from each other (i.e. normalizing in reciprocal circuits)
  • buffer lenghts longer than ~2k(5k) significantly slow-down script
    • likely computing xcorr not rendering it
  • noise variance only matters indirectly
    • likely

Simple:

  • delay has predictable signature on xcorr
    • (if autocorr is minimal, influence is delta function)

methods requirements

interactivity requirements

  • editable params

    • internal noise per node
    • weights, delay per edge
    • could visualize weights and delay as a synaptic waveform ?
  • could have GUI simply be draggable points

    • with contrained limits, background squares
  • processing GUI libraries

    • p5.touchgui
      • last updated 2019
      • has some nice elements (slider2d, joystick)
    • p5.gui
      • recent
      • magic constructors -> likely fast to set up
      • no 2D slider?
    • (x) DOESNT actually work with p5 it seems
    • ControlP5 classic, clean
      • lots of elements (including slider2d)
      • dropdowns, radial dials
  • javascript GUI libraries

  • see also graphing libraries

  • graphica

    • demos visualization requirements
  • visual reflection of implicit params like weight, delay

  • color of output signal blends colors of inputs!

  • might want some fading trail of previously chosen params

    • or visual indication of what signal looks like with slider val +/- a bit
  • simple equation rendered as latex

import requirements

  • import graph as from textfield

  • render graph as circles

  • edit connections via drag and drop

  • plot 1D signals plotWave()

  • colored latex from dynamics-visualizer-p5


project sidegoals

  • could be an opportunity to use a library OTHER THAN p5, particularly for plotting
    • some mix of p5 for interaction, plotly for plotting?