Minor doc improve #985
Minor doc improve #985
Conversation
aminya
force-pushed
the
docImprove
branch
2 times, most recently
from
616686e
to
0b8a2ff
Compare
Co-Authored-By: Mihir Gadgil <scimas@users.noreply.github.com>
Co-Authored-By: Mike J Innes <mike.j.innes@gmail.com>
|
Could you merge this? Doc improvement PRs should be in bit size, so they get merged fast. I will create more PRs, as I read more |
You are right. It's still possible to split that PR up, though! |
|
What is holding this back from getting merged? 😞 People will stop contributing if their PRs don't merge. |
| @@ -20,20 +20,21 @@ julia> d2f(2) | |||
| 6 | |||
| ``` | |||
|
|
|||
| When a function has many parameters, we can get gradients of each one at the same time: | |||
| When a function has many parameters, we can get gradients of each one at the same time by passing each one as an argument: | |||
There was a problem hiding this comment.
I would keep only this change and leave the rest as it is since it seems redundant
There was a problem hiding this comment.
What changes are redundant? Do you mean the information about the example!?
There was a problem hiding this comment.
all changes except the line above, in my opinion. I prefer the previous version
There was a problem hiding this comment.
I am describing the syntax. I don't dee this as redundant.
There was a problem hiding this comment.
Apologies for letting this fall through the cracks. I've added some suggestions to deduplicate the wording and slim down the copy a bit. @CarloLucibello do you mind taking another look?
| @@ -20,20 +20,21 @@ julia> d2f(2) | |||
| 6 | |||
| ``` | |||
|
|
|||
| When a function has many parameters, we can get gradients of each one at the same time: | |||
| When a function has many parameters, we can get gradients of each one at the same time by passing each one as an argument: | |||
There was a problem hiding this comment.
| When a function has many parameters, we can get gradients of each one at the same time by passing each one as an argument: | |
| When a function has many parameters, we can get the gradient with respect to each in a single call: |
|
|
||
| ```jldoctest basics | ||
| julia> f(x, y) = sum((x .- y).^2); | ||
|
|
||
| ``` | ||
| For this `f`, we can get the gradient with respect to both `x` and `y` in a single call: |
There was a problem hiding this comment.
| For this `f`, we can get the gradient with respect to both `x` and `y` in a single call: |
If the "with respect to" is moved up top, this line is no longer necessary.
| julia> gradient(f, [2, 1], [2, 0]) | ||
| ([0, 2], [0, -2]) | ||
| ``` | ||
| where `x = [2, 1]` and `y = [2, 0]`. The above syntax is the same as: `df(x,y) = gradient(f,x,y); df([2,1], [2,0])` |
There was a problem hiding this comment.
| where `x = [2, 1]` and `y = [2, 0]`. The above syntax is the same as: `df(x,y) = gradient(f,x,y); df([2,1], [2,0])` | |
| The above syntax is the same as: `df(x,y) = gradient(f,x,y); df([2,1], [2,0])` |
There was a problem hiding this comment.
Having a statement like "where ..." is a little cumbersome. Since df(x, y) already names the parameters, there's no need to define them in prose. Alternatively, you could actually declare them in the doctest with x, y = ..., but that might be confusing given f(x, y) has already been defined earlier (users who aren't comfortable with scoping may conflate the two).
|
Feel free to update the branch. |
Ready
Doesn't overlap with #853