In the lecture videos in week 2 of class 5, ‘Word2Vec’ a new parameter is introduced, Theta subscript t, however the exact meaning of this term is not clear to me. The lecture describes it as “The parameters associated with output t” or “what is the chance of a particular word t being the label”.
Is this the column vector for element t of the Embedding matrix?
If not, could you elaborate more on what this is?
Thanks, David
Hey DTodd,
The subscripts identify one element within a larger object and so, the subscripts on (t) are trying to select one particular element from the given number of words at one particular position.
And so, what are you actually doing is:
There’s a link that describes all the superscripts and subscripts used in the DLS courses from the beginning till the end and I have found it useful. It may not directly answer your query but would simplify the typical variables used.
Thanks Rashmi.
And I’ll take a look at the link you included.
I have the same question and I did not understand the answer.
So, what is Theta? Is it another matrix (which is different from E) or a vector?
Is Theta trainable? If yes, then at which point it gets trained (changed)?
In most machine learning literature, the weights are known as “theta”, rather than the ‘w’ used in this course.
The two terms are interchangeable.
For those coming here, I’ve tried to get a more detailed follow-up on a related thread:
