Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
mastodon.gamedev.place is one of the many independent Mastodon servers you can use to participate in the fediverse.
Mastodon server focused on game development and related topics.

Administered by:

Server stats:

5.4K
active users

mastodon.gamedev.place: AboutProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.7

Public
Rune Skovbo Johansen

I'm working on a tool that attempts to tweak the parameters of my creature parametrization to match the look of a reference 3D model. It does this solely by looking at the silhouettes and executing the parameter changes which make the silhouettes match more closely. 🧵

Rune Skovbo Johansen

If anyone know of popular proven techniques for dynamically adjusting the step size in gradient descent, or how to handle approaching a minimum (switch to binary search?), or even escaping local minima, let me know!

Sep 11, 2024, 07:02 PM·Public·Phanpy
1boost·0favorites
Public
Avi Bryant

@runevision look at ADAM for step sizes. You could also try L-BFGS (not certain how it will do on 106 params). Can try random restarts to avoid local minima.

Quiet public
Rune Skovbo Johansen

@avi I'm looking into ADAM but all descriptions I could find assume familiarity with stochastic gradient descent. And when I'm looking into that, I can't figure out how to map the terminology to what I'm doing, or if it even applies. In the image here (from Wikipedia) I don't know what the summand Qi functions are supposed to correspond to. I have many parameters/dimensions (104) but I only have one evaluation function. I also don't know if I have anything corresponding to "i-th observation".

Text from Wikipedia: Both statistical estimation and machine learning consider the problem of minimizing an objective function that has the form of a sum: Q ( w ) = 1 n ∑ i = 1 n Q i ( w ) , where the parameter w that minimizes Q ( w ) is to be estimated. Each summand function Q i is typically associated with the i-th observation in the data set (used for training).
Quiet public
Avi Bryant

@runevision I think you can just treat yours as a degenerate case where you have exactly 1 observation, at which point SGD = GD. The momentum and adaptive learning rates from Adam should still apply.

Quiet public
Avi Bryant

@runevision alternatively, and maybe more conventionally, you could consider each pixel distance to be a separate observation.

ExploreLive feeds
About