In Joone there isn't a perceptron, like that one
described by the literature. The LinearLayer is not the
implementation of a perceptron, because it doesn't
learn the bias values.
In fact, a real linear layer (like a perceptron) should
have a bias and be able to adjust that bias in the
learning process. It should just sum up its inputs
(equal to beta=1 in LinearLayer) TOGETHER with its
bias. At the moment, the
linear layer doesn't have a bias.
Currently, in Joone the linear layer is useful as a
buffer, like we have written in the manual. However, it
is not a real linear layer in the sense of a layer that
sums up its inputs, together with its bias and able to