I am currently playing around with actor-critic methods. In my current problem, the actor has four outputs: pol = Chain(Dense(6, 32, relu), Dense(32, 4)) However, depending on the state only a subset of the actions are legal. So for choosing an action I use: x = get_featurevector(s, pos_a) probs …

Updating only a subset of parameters in a neural network using Flux

trahflow November 30, 2021, 5:59pm 2

Just multiplying the output of the actor with a boolean mask is not an option?

pol(x) .* (1:4 .∈ Ref(pos_a))

Topic		Replies	Views
Is it possible to selectively update the params in Flux.jl? Machine Learning flux	9	968	April 17, 2019
Flux.jl: Changing the value of activations from hidden nodes during backprop Machine Learning flux	0	525	March 30, 2019
Network Not Updating, Flux Julia Machine Learning	6	405	September 19, 2022
How to manually update the params in a `Flux.Chain` neural network? General Usage flux , machine-learning	1	988	July 19, 2020
How to enforce weight matrix to be of a certain form (train a subset) in Flux? Machine Learning question , flux	5	809	February 26, 2021