# workshop_script.R
# Peter Bubenik
# January 2017
# script for computer workshop: Topological Data Analysis with R
#
# Data: triangulated 3d figures in nonrigid3d.zip from 
# http://tosca.cs.technion.ac.il/book/resources_data.html
# We filter these triangulations and calculate persistent homology using 
# Perseus from http://www.sas.upenn.edu/~vnanda/perseus/
# We calculate persistence landscapes using the Persistence Landscapes Toolbox from
# https://www.math.upenn.edu/~dlotko/persistenceLandscape.html
# We compare figure types pairwise and perform (multiclass) classification.

# Here we set some user-defined parameters - you can change these

#main_directory <- "C:/Users/Leo/Desktop/R/workshop" # sample directory for windows users
main_directory <- "~/R/workshop" # sample directory for Mac/linux users

operating_system <- "Mac" # PC Mac or Linux
perseus_directory <- "perseus_output"
pl_directory <- "pl_output"
animation_directory <- "animation"
binary_directory <- "bin"
data_directory <- "nonrigid3d"

grid_size <- 200 # grid used for Perseus and Persistence Landscape Toolbox
max_depth <- 200 # max number of persistence landscape functions
max_filtration_value <- grid_size
pl_param_vals <- seq(0,grid_size,0.5) # values at which to evaluate the persistence landscape

max_pose_num <- 1 # 50 # maximum pose number for one figure, starts at 0
#figures_to_use <- c("cat","centaur","david","dog","gorilla","horse","lioness","michael","seahorse","shark","victoria","wolf") # cat,centaur,david,dog,gorilla,horse,lioness,michael,seahorse,shark,victoria,wolf
figures_to_use <- c("cat","dog")
figures_to_compare <- c("cat","dog")
#num_frames <- 100 # number of frames in animation of filtration
#animation_delay <- 5 # delay time between frames
N <- 100 # number of repeats in permutation test
cost <- 10 # lower: softer margin, higher: harder margin, default=1
num_folds <- 10 # number of folds for cross validation
show_individual_figures <- TRUE # draw filtered simplicial complexes, persistence diagrams, persistence landscapes
pca_coords_to_use <- 1:10 # number of pca coordinates to use for classification using pca
degrees_to_use <- 0:1 # which homological degrees are used for classification

# now the main part

# install the necessary packages if not yet installed
list.of.packages <- c('scatterplot3d','rgl','FNN','e1071','kernlab')
new.packages <- list.of.packages[!(list.of.packages %in% installed.packages()[,"Package"])]
if(length(new.packages)) install.packages(new.packages)

library(scatterplot3d) # needed for 3d plotting
library(rgl) # needed for 3d graphics
library(FNN) # needed for k-nearest neighbor
library(e1071) # needed for PCA, SVM
library(kernlab) # needed for SVM, kernel SVM

# set up the needed directories and files
setwd(main_directory) # change to the main directory
dir.create(file.path(main_directory, perseus_directory)) # comment this line after running once
dir.create(file.path(main_directory, pl_directory)) # comment this line after running once
if (!file.exists(file.path(main_directory,pl_directory,"configure")))
  file.copy(file.path(main_directory,binary_directory,"configure"),file.path(main_directory,pl_directory)) # PlotOfLandscapes needs the configure file to be in its working directory

source("tda_functions.R") # load Peter Bubenik's topological data analysis functions.

# get list of figures
figures <- vector() # initialize vector
for (figure_name in figures_to_use) # for each type of figure
  for (i in 0:max_pose_num)
    if (file.exists(paste(file.path(data_directory,figure_name),i,".vert",sep="")))
      figures <- rbind(figures, c(figure_name,i))
num_figures <- dim(figures)[1]

source("persistence_script.R") # load, filter, visualize, persistence landscapes saved to text files
load("nonrigid3d_landscape.RData") # use this after running persistence_script.R once

# Pairwise Comparison of figures_to_compare
vector_length <- length(pl_param_vals) # use for plotting 
pl_comparison(plm_list, figures[,1], vector_length, figures_to_compare, N, cost, degrees=0:1, 
              figures_to_use) # pval=

# Classification of figures_to_use
if (num_figures < length(pca_coords_to_use))
  pca_coords_to_use <- 1:3
pl_classification(plm_list, figures[,1], vector_length, figures_to_use, degrees_to_use, 
                  num_folds, cost, pca_coords_to_use)