code_data

CODE_DATA transforms between input data and formatted spike times

Description:
Function to return an input spike time cell array or desired output spike
time cell array depending on Type.  The type of coding performed on the 
inputs is determined by the Coding structure and is user expandable.

Syntax:
CodedData=CODE_DATA(PutNeurons, Type, Puts, Coding);

Input Parameters:
o PutNeurons: Array of input or output neuron indices, depending on
    whether inputs or outputs are being coded.
o Type: String with possible values 'Encode' or 'Decode' indicating
    whether Puts should be encoded into input spike times or decoded into
    desired output spike times.
o Puts: System inputs or outputs before encoding into input spike times or
    before decoding from output spike times. For Spike coding, the format
    is a cell array, with each cell containing a input or output pattern.
    Each pattern is a qx2 array, where the first column ranges from 1 up 
    to the number of input or output neurons, and the second column is the
    actual input or outputs. For linear coding, the format is n by m array
    where n is the number of input patterns and m is the number of input
    or output neurons
o Coding: A structure containing fields specific for the type of encoding
    or decoding to be performed. All methods require EncodeMethod or
    DecodeMethod. Linear coding requires InputRange and InputSpikeRange or
    OutputRange and OutputSpikeRange.

Output Parameters:
o CodedData: Formatted input or desired output spike times ready for use
    in other functions, such as train.

Examples:
>> ThNN=theta_neuron_network;
>> Coding.EncodeMethod='Spikes';
>> Coding.DecodeMethod='Spikes';
>> InputSpikeTimes=code_data(ThNN,'Encode',{[1 2], [1 4]},Coding)
>> DesiredSpikeTimes=code_data(ThNN,'Decode',{[1 23], [1 20]},Coding)

>> ThNN=theta_neuron_network('StructureFormat',{'LayerArray',[2 3 2]});
>> Coding=struct(...
      'EncodeMethod','Linear',...
      'DecodeMethod','Linear',...
      'InputSpikeRange',[2 8],...
      'OutputSpikeRange',[20 30],...
      'InputRange',[0 1],...
      'OutputRange',[0 1]);       
>> Inputs=rand(5,2); %5 patterns and 2 inputs (not including the ref input)
>> InputSpikeTimes=code_data(ThNN,'Encode',Inputs,Coding)
>> DesiredSpikeTimes=code_data(ThNN,'Decode',Inputs.^2,Coding)

See also theta_neuron_network

0001 function CodedData=code_data(PutNeurons, Type, Puts, Coding)
0002 %CODE_DATA transforms between input data and formatted spike times
0003 %
0004 %Description:
0005 %Function to return an input spike time cell array or desired output spike
0006 %time cell array depending on Type.  The type of coding performed on the
0007 %inputs is determined by the Coding structure and is user expandable.
0008 %
0009 %Syntax:
0010 %CodedData=CODE_DATA(PutNeurons, Type, Puts, Coding);
0011 %
0012 %Input Parameters:
0013 %o PutNeurons: Array of input or output neuron indices, depending on
0014 %    whether inputs or outputs are being coded.
0015 %o Type: String with possible values 'Encode' or 'Decode' indicating
0016 %    whether Puts should be encoded into input spike times or decoded into
0017 %    desired output spike times.
0018 %o Puts: System inputs or outputs before encoding into input spike times or
0019 %    before decoding from output spike times. For Spike coding, the format
0020 %    is a cell array, with each cell containing a input or output pattern.
0021 %    Each pattern is a qx2 array, where the first column ranges from 1 up
0022 %    to the number of input or output neurons, and the second column is the
0023 %    actual input or outputs. For linear coding, the format is n by m array
0024 %    where n is the number of input patterns and m is the number of input
0025 %    or output neurons
0026 %o Coding: A structure containing fields specific for the type of encoding
0027 %    or decoding to be performed. All methods require EncodeMethod or
0028 %    DecodeMethod. Linear coding requires InputRange and InputSpikeRange or
0029 %    OutputRange and OutputSpikeRange.
0030 %
0031 %Output Parameters:
0032 %o CodedData: Formatted input or desired output spike times ready for use
0033 %    in other functions, such as train.
0034 %
0035 %Examples:
0036 %>> ThNN=theta_neuron_network;
0037 %>> Coding.EncodeMethod='Spikes';
0038 %>> Coding.DecodeMethod='Spikes';
0039 %>> InputSpikeTimes=code_data(ThNN,'Encode',{[1 2], [1 4]},Coding)
0040 %>> DesiredSpikeTimes=code_data(ThNN,'Decode',{[1 23], [1 20]},Coding)
0041 %
0042 %>> ThNN=theta_neuron_network('StructureFormat',{'LayerArray',[2 3 2]});
0043 %>> Coding=struct(...
0044 %      'EncodeMethod','Linear',...
0045 %      'DecodeMethod','Linear',...
0046 %      'InputSpikeRange',[2 8],...
0047 %      'OutputSpikeRange',[20 30],...
0048 %      'InputRange',[0 1],...
0049 %      'OutputRange',[0 1]);
0050 %>> Inputs=rand(5,2); %5 patterns and 2 inputs (not including the ref input)
0051 %>> InputSpikeTimes=code_data(ThNN,'Encode',Inputs,Coding)
0052 %>> DesiredSpikeTimes=code_data(ThNN,'Decode',Inputs.^2,Coding)
0053 %
0054 %See also theta_neuron_network
0055 
0056 %Copyright (C) 2008 Sam McKennoch <Samuel.McKennoch@loria.fr>
0057 
0058 
0059 if strcmpi(Type,'EncodeInputs')
0060     Inputs=Puts;
0061     InputNeurons=PutNeurons;%get_input_neurons(ThNN);
0062     
0063     switch lower(Coding.EncodeMethod)
0064         case ('spikes')
0065             CodedData=cell(1,length(Inputs));
0066             for j=1:length(Inputs) %For multiple patterns
0067                 for k=1:size(Inputs{j},1) %If there are multiple inputs
0068                     CodedData{j}(k,:)=[InputNeurons(Inputs{j}(k,1)) Inputs{j}(k,2)];
0069                 end
0070             end
0071             
0072         case ('linear')
0073             NormalInputs=(Inputs-Coding.InputRange(1))/(Coding.InputRange(2)-Coding.InputRange(1));
0074             tiInputs=NormalInputs*(Coding.InputSpikeRange(2)-Coding.InputSpikeRange(1))+Coding.InputSpikeRange(1);
0075             
0076             if length(InputNeurons)<size(tiInputs,2)
0077                 disp('Error in code_date: Not enough input neurons');
0078                 CodedData=-1;
0079                 return;
0080             end
0081             
0082             CodedData=cell(1,size(tiInputs,2));
0083             for j=1:size(tiInputs,1) %For multiple patterns
0084                 for k=1:size(tiInputs,2) %If there are multiple inputs
0085                     CodedData{j}(k,:)=[InputNeurons(k) tiInputs(j,k)];
0086                 end
0087             end
0088 
0089         %Bad coding method
0090         otherwise
0091             disp('Error: Invalid Encoding Method in code_data function');
0092             return;
0093     end
0094 
0095 
0096 elseif strcmpi(Type,'DecodeOutputs')
0097     Outputs=Puts;
0098     OutputNeurons=PutNeurons;%get_output_neurons(ThNN);
0099     
0100     switch lower(Coding.DecodeMethod)
0101         case ('spikes')
0102             CodedData=cell(1,length(Outputs));
0103             for j=1:length(Outputs) %For multiple patterns
0104                 for k=1:size(Outputs{j},1) %If there are multiple inputs
0105                     CodedData{j}(k,:)=[OutputNeurons(Outputs{j}(k,1)) Outputs{j}(k,2)];
0106                 end
0107             end                        
0108         
0109         case ('linear')
0110             NormalOutputs=(Outputs-Coding.OutputRange(1))/(Coding.OutputRange(2)-Coding.OutputRange(1));
0111             tdOutputs=NormalOutputs*(Coding.OutputSpikeRange(2)-Coding.OutputSpikeRange(1))+Coding.OutputSpikeRange(1);
0112             if length(OutputNeurons)<size(tdOutputs,2)
0113                 disp('Error in code_date: Not enough output neurons');
0114                 CodedData=-1;
0115                 return;
0116             end
0117             
0118             CodedData=cell(1,length(tdOutputs));            
0119             for j=1:size(tdOutputs,1) %For multiple patterns
0120                 for k=1:size(tdOutputs,2) %If there are multiple inputs
0121                     CodedData{j}(k,:)=[OutputNeurons(k) tdOutputs(j,k)];
0122                 end
0123             end
0124 
0125         %Bad coding method
0126         otherwise
0127             disp('Invalid Decoding Method in code_spike_times function!');
0128     end
0129     
0130 elseif strcmpi(Type,'EncodeOutputs')
0131     %Change output spikes into dataset outputs
0132     OutputSpikes=Puts;
0133     OutputNeurons=PutNeurons;%get_output_neurons(ThNN);
0134     
0135     switch lower(Coding.DecodeMethod)
0136          case ('spikes')
0137             for j=1:length(OutputSpikes) %For multiple patterns
0138                 for k=1:length(OutputNeurons) %If there are multiple outputs
0139                     CodedData{j}(k,:)=[k OutputSpikes{j}(k,2)];
0140                 end
0141             end
0142 
0143          case ('linear') %Assumes SISO
0144             for j=1:length(OutputSpikes) %For multiple patterns
0145                 for k=1:length(OutputNeurons) %If there are multiple outputs
0146                     Temp=(OutputSpikes{j}{OutputNeurons(k)}-Coding.OutputSpikeRange(1))/(Coding.OutputSpikeRange(2)-Coding.OutputSpikeRange(1));
0147                     CodedData(j,k)=Temp*(Coding.OutputRange(2)-Coding.OutputRange(1))+Coding.OutputRange(1);
0148                 end
0149             end
0150 
0151         %Bad coding method
0152         otherwise
0153             disp('Invalid Decoding Method in code_spike_times function!');
0154     end
0155 elseif strcmpi(Type,'DecodeInputs')
0156     %Change input spikes into dataset inputs
0157     InputSpikes=Puts;
0158     InputNeurons=PutNeurons;
0159     switch lower(Coding.EncodeMethod)
0160         case ('spikes')
0161             CodedData=cell(1,length(InputSpikes));
0162             for j=1:length(InputSpikes) %For multiple patterns
0163                 for k=1:size(InputSpikes{j},1) %If there are multiple inputs
0164                     CodedData{j}(k,:)=[k InputSpikes{j}(k,2)];
0165                 end
0166             end
0167         case ('linear') %Assumes SISO
0168             for j=1:length(InputSpikes) %For multiple patterns
0169                 for k=1:size(InputSpikes{j},1) %If there are multiple inputs, use this instead of InputNeurons to avoid figuring out ReferenceTime here
0170                     Temp=(InputSpikes{j}(k,2)-Coding.InputSpikeRange(1))/(Coding.InputSpikeRange(2)-Coding.InputSpikeRange(1));
0171                     CodedData(j,k)=Temp*(Coding.InputRange(2)-Coding.InputRange(1))+Coding.InputRange(1);
0172                 end
0173             end
0174 
0175         %Bad coding method
0176         otherwise
0177             disp('Invalid Decoding Method in code_spike_times function!');
0178     end    
0179 else
0180     disp('Error: Invalid Coding Type in code_data function');
0181     SpikeTimes=-1;
0182     return;
0183 end