%PS2 Solution
%1.  Floor data using smooth function, take log
%2.  Compute normalization constants for ML and MAP estimation.  Normalize
%		data using the following norm methods:
%			Max Likelihood
%			Max a Posteriori
%			Whole chip
%			Arithmetic mean
%			No normalization




%Call the dataset being normalized unnorm_data.  That way you
%only have to change the name in one place to go from normalizing
%the test data to normalizing the real data.
unnorm_data = test;

%control_rows is the number of rows containing "controls".
%For test, it's 5, for PS2 it's 15.
control_rows = 5;

%get the total number of rows being normalized, find missing data
total_rows = size(unnorm_data,1);
missing_data = unnorm_data==-1000;

%1.  Floor data using a smooth function such that values less than 40 
%    (including negativevalues) are all compacted into the range 
%    between 0 and 40.  Use the following transform:
%       x --> x                         x in  (40,INF)
%       x --> (x^2)/160 + x/2 + 10      x in  [-35,40]
%       x --> 0.15625                   x in (-INF,-35)
%
%	Then, take the log.


%!YOUR CODE HERE, you should end up with 2 new data sets, log_floor_data and floor_data


%2.  Compute Max Likelihood and MAP of normalization parameters
%		Using functions that you have written (added code to)
%		ML_norm takes as input the unnormalized (log) controls,
%			and returns the normalization (log) constants
%		MAP_norm takes as input the unnormalized (log) controls,
%			and two constants, alpha and t.
%			Those are currently set as 3 and 1, you can change them if you like
%			and see what the effects are.
%			MAP_norm also returns the normalization (log) constants.
%
%calculate normalization constants for the ML and MAP methods
ML_norm_consts = ML_norm(log_floor_data(1:control_rows,:));
MAP_norm_consts = MAP_norm(log_floor_data(1:control_rows,:),3,1);
%normalize the data
norm_ML = log_floor_data-repmat(ML_norm_consts,total_rows,1);
norm_ML(missing_data) = -1000;
norm_MAP = log_floor_data-repmat(MAP_norm_consts,total_rows,1);
norm_MAP(missing_data) = -1000;
norm_WC = log(WC_normalize(floor_data, missing_data,WC_data));
norm_AM = log(AM_normalize(floor_data, missing_data,control_rows));