# AP_1_ex_2
K-nearest-neighbors: a server/client implementation (KNN_ClientServer, client)

The server is the KNN classifier. It finds the type(classification) of an input vector by comparing it
with a data set of classified vectors using the K-nearest-neighbors algorithm.

The client receives input from user, send to server, receives classification back from server,
and prints to screen.

The server's program takes the following arguments in order:
1. name of file containing a set of classified vectors formatted as follows:
    > the values of the vectors must be formatted such that they can be interpreted as floating-point numbers.
    > this can be in standard or scientific notation (the latter only with a capital "E").
    > each of the values must be followed by a comma (",") and the last value must be followed by a type name.
    > each of the vectors must be on separate lines.
    > the vectors must all be the same size.
    > there must be no text other than properly fomatted vectors.
2. port number: must be between 0 and 65536.

The clients program takes the following arguments in order:
1. ip: server's ip address
2. port number: the number of the port the server in listening to

The client runs in an infinite loop and receives from user the following arguments in order:
1. vector which type you would like to find using the server's classifier.
   enter the vector as follows:
    - values formatted to be interpreted as floating point numbers (see file requirements above)
    - each value followed by a space
    - exactly the same number of values as the vectors in the given file
2. three letter distance method code. supported distance methods and their codes are as follows:
    a) "AUC" - Euclidean distance
    b) "MAN" - Manhattan distance
    c) "CHB" - Chebyshev distance
    d) "CAN" - Canberra distance
    e) "MIN" - Minkowski distance w/default p parameter of 2 (user defined p value is not yet supported)
3. K-value for the KNN algorithm - must be a positive integer

 
Program files:
    - server_main.cpp
    - KNN_ClientServer.h
    - KNN_ClientServer.cpp
    - KNN_Set.h
    - KNN_Set.cpp
    - ClassifiedVector.h
    - ClassifiedVector.cpp
    - VectorDist.h
    - VectorDist.cpp
    - makefile
    
Compile instructions:
1. open the directory containing the program files with the terminal.
2. check that directory contains all of the program files above.
3. enter 'make' command in the terminal to compile.

Program use instructions:
1. in the terminal, open the directory containing the compiled executable "server.out", "client.out".
2. to start the program enter the following command:

    "./server.out <data set filename> <port number>"

    then open another terminal and enter:

    "./client.out <ip address> <port number>"
    
    - for example "./server.out iris_classified.csv 12345"
                  "./client.out 127.0.0.1 12345"

if the arguments and file given are valid, both programs will initialize.
note that the server will compile the file into a dataset of classified vectors.

now you may enter into the client's terminal the arguments discussed above
in order to find a vector's classification, with your chosen distance metric and k neighbors.
    
To find the type of a vector using the given data set, distance function, & k-value using
the KNN algorithm, enter the vector into the terminal as described above.

The program will then parse the given vector, the distance metric, and the number of neighbors,
and print the vector's type found using the KNN algorithm (run by the server) with the given parameters.

4. after the type is printed, the client's program will take another input.
If an invalid input is given, the client's program will print "invalid input" and wait for another input.
This will continue indefinitely until "-1" is given, which will close the connection.

5. the server, once initialized, will always run and accept clients. If a connection with a client is formed,
the server will handle this client only until it decides to terminate the connection (using "-1" as input).
At this point the server will accept new clients, if exist, or wait indefinitely for a client to form a connection.