Summary
I have added the option to randomly select initial trees from .t file. The result will be the best one among all results. 
I have modified many classes and methods. So it's quite different from previous version. I have added a class resultSet to manage all information about results.(Because now there are multiple results to select) Now there is only one output file. The information about stop criteria has been included in this file.


Command and Options:
to compile: javac centroid.java
to run: java centroid option1=value1 option2=value2 ...

example:
java centroid tfile=seq3.nex1.run1.t randomnumber=5 hastruetree=1 itr=1000 algorithm=3 truetreefile=seq3.tree inittreetype=5

options:

help			 show help information
tfile                    the .t file
pfile                    the .p file
truetreefile             the file contains the true tree
inittreefile             the file contains the initial tree
itr                      the number of iterations
algorithm                1:K-Interval; 2:Quartet; 3:SD; 4:K-Interval Sqr
inittreetype             1: the tree with biggest LNL value in .p file
                         2: the most frquently happened tree in .t file
                         3: neighbor join tree
                         4: phyml tree
                         5: random trees from .t file
randomnumber             the number of random trees
hastruetree              0: does not have true tree; 1: has true tree
stricthillclimbing       0: involves prob, c, temp; 1: use stricthillclimbings


Class and Methods:

node:		the class for leaf node object

edge:		the class for edge object

resultSet:	the class for manage result information. It contains: the initial tree, the result tree, the stop criteria, the e_T value for the result tree, the c value if used, the number of moves happened.


tree:		the class for implement tree object. Its methods can be divided into two parts.

format converter: these methods is used to convert the tree between tree object and Newick string (both rooted and unrooted, leaf as taxa or id)

int NewiToTree(String s, int pointer, node currentnode, SortedMap<String,String> map)

String TreeToNewi(node r, SortedMap<String,String> map)

String getUnrootedTree(node r, SortedMap<String,String> map)

// the map in these methods is used to add taxa for leaves, instead of leaf ids


data formalizer: there methods is used to formalize the tree and its data (k-interval, SD, quartet, NNI)

String NNI(edge e, int lowerid, int upperid)

ArrayList<String> NNIlist()

void setMACAPair(node r)

void setQuartetSet()

void setKInterval()

void setEdgePartition()

// setEdgePartition() is used for SD algorithm


centroid:	the main class which implements the algorithm

Some important public variables:

    //initTreeType sets the type of the initial tree.
    // 1: the sample tree in .p file with biggest LNL 
    // 2: the sample tree in .t file which has been repeated for most times
    // 3: the neighbor joining tree
    // 4: the phyml tree
    // 5. random initial trees taken from .t file
    public static int initTreeType = 1;

    // hasTrueTree decides if a true tree is provided
    public static boolean hasTrueTree = false;

    // strictHillClimbing decides if the algorithm uses strict hill climbing or temp, c, prob
    public static boolean strictHillClimbing = true;

    // algorithm sets the algorithm to use
    // 1: k-interval
    // 2: quartet
    // 3. SD
    // 4. k-interval square
    public static int algorithm = 1;

    //the size of the samples
    public static int size = 0;

    // the number of sample trees
    public static int tree_num = 0;

    // the pre-calculated value for k-interval
    public static long N_c_sqr = 0;
    public static ArrayList<Integer> N_c = new ArrayList<Integer>();

    // the pre-calculated value for SD
    public static SortedMap<String, Integer> SD_set = new TreeMap<String, Integer>();

    // stop_criterion sets the step criterion. Not used in current version. Now all criterions are in use.
    public static int stop_criterion = 0;

    // the list which contains all trees in newick format from .t file
    public static ArrayList<String> newickTreeList = new ArrayList<String>();

    // the map between id and taxa. They are used to translate leaves between ids and taxas
    public static SortedMap<String,String> nodeMap = new TreeMap<String,String>();
    public static SortedMap<String,String> node2idMap = new TreeMap<String,String>();

    // divideBy decides by how much the c value will be divided
    public static int divideBy = 1;

    // the true tree
    public static tree TrueTree = new tree();

    // the result list
    public static ArrayList<resultSet> result_list = new ArrayList<resultSet>();
    
    // the quartet tree list contains the puartet data for all trees
    public static ArrayList<ArrayList<Integer>> quartetTreeList = new ArrayList<ArrayList<Integer>>();
    
    // the number of the initial tree
    public static int init_tree_num = 0;


Important methods:

ArrayList readFile(String file)

void getInitTree(String p_file, String t_file, String mltree)

int getDistance(tree t1, tree t2)

long getE_T(tree t, ArrayList<ArrayList<Integer>> quartetList)

long setC(tree t, ArrayList quartetList, int algorithm)

tree centroid(int itr, long c, resultSet res, ArrayList quartetList)

resultSet hillclimbing(String p_file, String t_file, String truetree, String mltree, int itr)

void summarize(resultSet res, ArrayList<ArrayList<Integer>> quartetList)

// centroid(int itr, long c, resultSet res, ArrayList quartetList) is the most important method