Writing a target vector in code for a pattern recognition model I am trying to build a pattern recognition neural network model.  
I was told that the target for pattern1 is 1, pattern2 is 0, and pattern3 is -1.  However, I am not sure how to create my weight, bias and target vectors?
My input vectors are the following:
pattern_1 = [1 1 1 -1];
pattern_2 = [1 -1 -1 1];
pattern_3 = [1 -1 1 -1];

How do I write the target vector for the patterns?  Is it:
weight = [0 0 0];
bias = [1 1 1];
target = [1 0 -1];

I will eventually want to perform:
a = weights * pattern_1
error = target - a

But I am unsure what sized matrix my target vector needs to be, given the fact that I was just told that it is either 1, 0 or -1.
 A: You can think of the bias as an additional feature whose value is 1.
Therefore with the bias, your input vectors are,
pattern_1 = [1 1 1 -1 1];
pattern_2 = [1 -1 -1 1 1];
pattern_3 = [1 -1 1 -1 1];

or you can write the input vectors as one matrix,
$$ X = 
\begin{bmatrix} 
1& 1  & 1&-1&1\\
1& -1  & -1&1&1\\
1& -1  & 1&-1&1
\end{bmatrix}
$$
where each row corresponds to each pattern vector.
Note that I added the bias values at the end of each input vector.
Your initial weight vector is [0 0 0 0 0] or a vector of random values. The weight vector should have a weight for every feature of your input vector.
Since the target values are in the set $\{-1,0,1\}$, it is a multi-classification problem. You might want to use the softmax objective function.
You can write the target values (denote it as $y$) as a matrix using one-hot encoding.
The target matrix is
$$ y = 
\begin{bmatrix} 
0& 0  & 1\\
0& 1  & 0\\
1& 0  & 0
\end{bmatrix}
$$
where each row corresponds to a target value of a pattern vector.
Index $j$ of a row $i$ of $y$ represents the target value of pattern (or sample) $i$. There are three indices for each row vector. Index 0 represents target -1, Index 1 target 0, and Index 2 target 1.
Now your dataset consists of the input matrix $X$ and the target matrix $y$. To optimise your weight vector $w$ you need to choose an objective function. You can start with the softmax objective function explained here.
