Multivariate Calculus
Differentials¶
is used to encode infinitesimal changes- used to act as a placegolder value
- divide wrt time to get rate of change
CHAIN RULE
Chain Rule with More Variables¶
Let
Gradient Vector¶
Note:
Directional Derivatives¶
Implications¶
Direction of
Lagrange Multipliers¶
Goal: minima/maximize a multi-variable function (
These can be obtained on combining the given restraints with the following.
Basic idea: to find
Note: Take care that the point is indeed a maxima or minima as required and not just a saddle point (second derivative test won't be applicable so be vigilant).
Functions | Example | Value | First derivative | Second derivative |
---|---|---|---|---|
loss function | ||||
neural net layer |
Gradient¶
Hessian¶
We have
Note: Hessians are square-symmetric matrices.
Jacobian¶
where
Examples¶
, if is symmetric , if is symmetric