Training Flashcards
1
Q
What is the Optimality Condition?
A
Each partial derivative of the objective function with respect to the weight parameters should be set to zero
2
Q
If a minimal error of zero can be achieved when training a linear model, then this is equivalent to solving which equations?
A
XW = Y
In other words, our prediction exactly matches the expected output.
3
Q
When there are more training samples than input features the system is …
A
Over-determined
4
Q
When we cannot directly calculate the inverse, what do we use to calculate the optimal weights
A
We use the Pseudo Inverse of the matrix representing the training data
