TOOLS is a collection of utilities for I/O, distance, similarity computation of discrete sequences and univariate time series.

Data Format
===========
Data is stored as text files. Each discrete sequence or time series is specified as a single line. A discrete sequence is represented as space separated string integers. Each integer represents a symbol. A time series is represented as space separated string of floating point values.

A sequence/time series is internally stored as a vector of integers/floats.
List of functions:

I/O
====
1. Tokenize - Parses a string into a vector of strings, integers or floats.
2. readSequences - Reads a vector of discrete sequences or time series from an input stream.
3. printSequences - Print sequences to an output stream.
4. printSequence - Print a sequence to an output stream.
Manipulation
============
1. loadMap - Load the map representing pairwise symbol weights (used by WSMC similarity measure).
2. printMap - Save the map representing pairwise symbol weights (used by WSMC similarity measure).
3. findMax - Find the maximum occurring symbol in a set of discrete sequences.


Discrete Sequence Similarity
============================
1. Simple Matching Coefficient
float SMC(std::vector<int> &a, std::vector<int> &b);
2. Weighted Simple Matching Coefficient
float WSMC(std::vector<int> &a,std::vector<int> &b,std::map<pair<int,int>,float>&alphaStd::Map);
3. Normalized Longest Common Subsequence using Dynamic Programming (O(mn))
float LCS_DP(std::vector<int> &a, std::vector<int> &b);
4. Normalized Longest Common Subsequence using Hybrid of Hunt Szymanski and Dynamic Programming (Budalakoti et al, 2007).
float LCS_HY(std::vector<int> &a, std::vector<int> &b,int max_symbol=-1,int max_occ_any_symbol=-1);
5. Distance between bitstd::maps (Kumar et al 2005)
float BITMAP(std::vector<int> &a,std::vector<int> &b, int level);
6. Similarity using cross correlation (Protopapas et al 2007)
float CROSSCORR(std::std::vector<float> &a, std::std::vector<float> &b,fftw_p
lan plan_forward, fftw_plan plan_backward, double *fin, fftw_complex *fout,fftw_complex *bin, double *bout);


Time Series Distance
====================
/* DIST - Wrapper function for calling different distance measures
   1 - Euclidean
   2 - Dynamic Time Warp (DTW)
   3 - DTW with Sakuro Chiba Band
   4 - LB_KEOGH (Keogh, E. (2002).  Exact indexing of dynamic time warping)
*/
float DIST(std::vector<float> &a,std::vector<float> &b,int measure);

float EUC(std::vector<float> &a,std::vector<float> &b);
float DTW(std::vector<float> &a,std::vector<float> &b);
float DTW_SC(std::vector<float> &a,std::vector<float> &b);
float LB_KEOGH(std::vector<float> &a, std::vector<float> &b);

/* pairwiseDist - Calculate pairwise distance between two sets of time series
 */
void pairwiseDist(std::vector<std::vector<float> > &sequences1,std::vector<std::vector<float> > &sequences2,std::vector<std::vector<float> > &pairwise, int msr);

/* queryNN - Finds the distance of query time series to the nn^{th} nearest neighbor in DB using the LBKeogh lower bound as proposed in Ratanamahantana and Keogh 2004. Returns the distance to the nn^th nearest neighbor and index stores its id in DB.
*/
float queryNN(std::vector<float> &query, std::vector<std::vector<float> > &DB, int nn, unsigned int *index);