Confusion on the algorithm used by DenseVariational

[pasq-cat]

Hi, i am learning probabilistic deep learning with tensorflow probability. i was able to fit my data by following the tutorial https://www.tensorflow.org/probability/examples/Probabilistic_Layers_Regression

but there is 1 information that i did not find. tfp.layers.DenseFlipout use the flipout estimator tfp.layers.DenseLocalReparameterization the local reparameterization trick as proposed by kingma 2015 ( sampling the activations) tfp.layers.DenseReparameterization the global reparameterization trick as proposed by kingma et al in 2014

tfp.layers.DenseVariational ????????????????????? what is the algorithm used ? it is the only one that does not report it.

[leandrolcampos]

Hi @Rockdeldiablo,

I think it's based on the method known as Bayes by Backprop (Blundell et al., 2015).

[pasq-cat]

Hi @Rockdeldiablo,

I think it's based on the method known as Bayes by Backprop (Blundell et al., 2015).

thanks