Master Algorithm

By Pedro Domingos, this focuses on the high-level algorithms, not on ethics (like http://doyoutrustthiscomputer.org). This pairs well with Rod Brook's post on AI Progress. The main idea off the book is breaking down different strategies of machine learning:

TribeOriginsAlgorithm
BayesiansStatisticsProbabilistic
AnalogizersPsychologyKernel
SymbolistsLogicInverse Deduction
EvolutionariesBiologyGenetic
ConnectionistsNeuroscienceBack Propogation

Comments:

  • Chapter 2 - The argument for evolution - saying ‘'evolution is an algorithm'' seems a little simplistic - from Sapiens it seems like a large number of random events have lead to the here and now.
  • Chapter 5 - From Naive Bayes to Hidden Markov chains to Kalman filters in 3 pages!
  • Chapter 7 - most recommendation systems are some sort of nearest-neighbor
    • Curse of dimensionality - Dimensions higher than 3 explained with pealing a hyperspace orange
    • what is a Support Vector Machine
  • Chapter 8
    • Putting together birds of a feather - several pages to describe k-means
    • Discovering the shape of the data - Principal Component Analysis - lining data up, reducing dimensions
    • Learning to Relate - What are Relational Learning algorithms?
  • Chapter 9
  • Chapter 10
  • Epilogue - Machine Learning examples