Description Usage Arguments Value Examples
This function combines one or more DeepBeliefNets, RestrictedBolzmannMachines and Layers into a single DeepBeliefNets.
The layers must be compatible in order, i.e. the output of the previous DBN/RBM must be the same (both in term of size and type) than the input of the next one.
The only exception is with Layer
objects, where RestrictedBolzmannMachine
s will be created before and after the layer
(only one RestrictedBolzmannMachine
is created Layer
succeeding immediately an other Layer
).
1 2 |
... |
objects to be combined |
biases.first |
whether to use the biases of RBM i (TRUE) or i+1 (FALSE) for shared layers. Defaults to TRUE as we don't care much about the Bs after the pre-training. |
the combined DBN. Note that it is not tagged as unrolled and fine-tuned any more. It is tagged as pre-trained if all individual DBNs/RBMs were pre-trained.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 | # DeepBeliefNets only
dbn1 <- DeepBeliefNet(Layers(c(784, 1000), input="continuous", output="binary"))
dbn2 <- DeepBeliefNet(Layers(c(1000, 500, 250), input="binary", output="binary"))
dbn <- c(dbn1, dbn2)
# RestrictedBolzmannMachines only
dbn <- c(dbn1[[1]], dbn2[[1]], dbn2[[2]])
# Layers only
dbn <- c(Layer(784, "continuous"), Layer(1000, "binary"), Layer(500, "gaussian"))
# Layers only
c(Layer(784, "continuous"), Layer(1000, "binary"), Layer(500, "binary"))
# Mixing it all
rbm3 <- RestrictedBolzmannMachine(Layer(250, "binary"), Layer(30, "binary"))
layer4 <- Layer(2, "gaussian")
c(Layer(2, "gaussian"), dbn1, dbn2, rbm3, layer4)
# The following won't work
## Not run:
dbn3 <- DeepBeliefNet(Layers(c(250, 500), input="binary", output="binary"))
dbn <- c(dbn1, dbn3)
dbn4 <- DeepBeliefNet(Layers(c(1000, 500), input="continuous", output="binary"))
dbn <- c(dbn1, dbn4)
## End(Not run)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.