hypothesis function [ML-CLASS]

[golirev]

Hi,

I've been doing the ML course from stanford on my own now that I have some time, and I cannot figure out why the course material shows the logistic regression hypothesis function as h = g(θ^T*x) where θ is a nx1 matrix (therefore θ^T is a is a 1xn) and X is a mxn matrix. Those dimensions are not compatible with each other. When I look at octave/matlab solutions the hypothesis function is implemented the other way around and without the transpose as h = g(x*θ). What am I missing? Why does this work?

Another way of making the dimensions compatible would be to transpose x but I haven't seen that done.

An explanation of the function is found on the following page
Machine Learning

Thanks

 

[BradtheRad]

Unaware how to do that type of math myself, but just trying to be helpful...

The explanation at your posted link appears to have been written for a different field than electronics. Whether or not that's the reason for the discrepancy, it is common in our schooling to be exposed to explanations that:

* are based on greater requisite knowledge than we have right now

* demonstrate a method primarily rather than an equation, in the intention that we will adapt the method to our own use

* were lifted from a treatise which contains a few missing details in another chapter, and that chapter was omitted from the article you're reading.

* are really too complicated to present in a linear manner, yet we can only read in a linear manner

* are about a topic that needs to be seen from different angles, to get a feel for the ways in which one branch joins to other branches, so to speak.

* are not entirely correct now, but may have been at an earlier time