2015.09.10: colormaps

catch up last time:

• Mai ggplot2. BB: see ggplot2 cheat sheet

Mai showed a A Better Default Colormap for Matplotlib | SciPy 2015 | Nathaniel Smith and Stéfan van der Walt - YouTube.

• default "jet" does poorly, eg doctors looking at patient data (Borkin et al 2011)

• Matlab switched default to "parula"

• 3 receptors so showing spectra, RGB, whitespace normalization

• CIE Lab (1976):

  Lab = matrix @ ((XYZ / XYZ_w) ** (1/3))

• CIECAM02, CAMO2-UCS: builds on estimating brightness and color distance

• color blindness

  pip install viscm
  python -m viscm view jet

• Demo: circular colormap

  N = 100

• colorblind and greyscale friendly: dark-blow-to-light-yellow

  pip install colorspacious
  python

• new "viridis", after a snake

• Matplotlib 2.0 = 1.0 + better styles

ColorBrewer: Color Advice for Maps

• qualitative

• sequential

• diverging

• BB: also Light & Bartlein (2004) The end of the rainbow? Color schemes for improved data graphics

• RColBrew.R

  library('RColorBrewer')
  display.brewer.all() # to display all color palettes
  
  cols = brewer.pal(4, 'BuPu')  # take 4 colors from the blue purple palette
  pal  = colorRampPalette(cols) # interpolate the 4 colors to return a function

• Stanford paper on qualitative

• iWantHue

• Practical Rules for Using Colors

• BB: can also write from R/Matlab/Python to PDF, and edit layout / elements in Adobe Illustrator

next time

• machine learning, like random forest, from Titanic Kaggle
  • introduce concepts, eg Andrew Ng's Machine Learning Course
• Titanic with Julia - Of Data Monsters