LSE

Intro:

Super Level Set Estimation Algorithms Implementation, including LSE, TRUVAR, RMILE and some improved version. Projects for stochastic process course in 2019 Spring.

Files:

• algos.py : classes of different algorithms
• utils.py : some useful functions, such as drawing plots
• main.py : main code for running experiments

Requirements:

• python=3.6
• numpy=1.17.0
• scipy=1.3.0
• matplotlib=2.0.2

Usage:

python main.py [--test_type TEST_TYPE] [--algo ALGO [ALGO ...]] [--cost COST]

• test_type : three different mode for test and drawing plots
  • normal (default) : run each algorithm for 10 times, calculate the avarage steps and draw F1 score plots with steps, saved in images/f1_step.png
  • cost : run each algorithm for 1 time, draw F1 score plots with costs, saved in images/f1_cost.png
  • single : run only one algorithm and draw first 20 picked points and paths, saved in images/algo_label+points.png, images/algo_label+paths.png
• algo : choosing which algo to run, need to input at least one algo, indicated by an integer number
  • 1 : LSE
  • 2 : LSE with implicit threshold
  • 3 : modified LSE with implicit threshhold (referenced in my project report)
  • 4 : TRUVAR
  • 5 : TRUVAR with implicit threshold
  • 6 : RMILE
  • 7 : LSE with considering cost (redundant, convenient for debugging)
• cost : whether to consider distance costs in algorithms
  • True : consider distance costs
  • False (default) : not consider distance costs