need more references for essay.md

[timm]

Here is what i want to students in my grad class to try:

• model readabilty
• learnability and repeatability of the results
  • i.e. using ess CPU and less disk/RAM
• mutli-goal reasoning
• anomaly detection (a little bell that goes "ping!" when new questions not in old data)
• incremental (knows how to forget old knowledge and learn new stuff, when required)
  • many use an anomaly detector
• sharable
  • can succinctly describe the training data (somehow)
  • and/or that data can be mutated to lower the odds of detecting idividuals within the data
  • and/or that summarized/mutated data still works well for learning a model
• context aware: generates different conclusions for different regions of the data
• self-tuning (knows how to tune its own parameters)

Here are the references I currently have :

• https://github.com/txt/fss17/blob/master/essay.md

[amritbhanu]

• importance on improving the data quality?
• dealing with big data, and how to run jobs in parallel such as in spark cluster?
• idea of transfer learning?

[rahlk]

Methods for streaming data

• Very fast decision trees (A classic): https://homes.cs.washington.edu/~pedrod/papers/kdd00.pdf
  • This is in weka
  • A number of magic parameters here. Perhaps we may tune them?
• Handling uncertainty in data with decision trees
  • See section 4.1 in this paper for simulating uncertainty. We may also use this for anomaly detection.
• Gama's simple incremental decision trees: Accurate decision trees for mining high-speed data streams