david on Nostr: grapevine math and neural network math are more similar than I had thought ...
grapevine math and neural network math are more similar than I had thought
Similarity: Both make use of a weighted sum.
Difference: Grapevine equation divides by the sum of weights (since it’s a weighted average), something you don’t do with neural networks.
Similarity: Neural networks multiply by the sigma function; the grapevine multiplies by a function that maps weights onto confidence. Both functions are nonlinear.
Difference: neural networks have a ‘bias’ parameter which does not show up in the grapevine equation.
Similarity: Both make use of a weighted sum.
Difference: Grapevine equation divides by the sum of weights (since it’s a weighted average), something you don’t do with neural networks.
Similarity: Neural networks multiply by the sigma function; the grapevine multiplies by a function that maps weights onto confidence. Both functions are nonlinear.
Difference: neural networks have a ‘bias’ parameter which does not show up in the grapevine equation.