RBM neural network predict using sklearn

Question

I'm attempting to use RBM neural network in sklearn, but I can't find a predict function, I see how you can train it (I think) but I can't seem to figure out how to actually predict a value.

http://scikit-learn.org/stable/auto_examples/neural_networks/plot_rbm_logistic_classification.html#example-neural-networks-plot-rbm-logistic-classification-py

I'm working on a class assignment. this is the assignment:

You will then use randomized hill climbing algorithm to find good weights for a neural network.

Is it possible to do this with SKLearn? Is there a better recommended tool to be able to select different weights for NN? (The goal is to experiment with around 3 different search optimization techniques, and learn about them, not necessarily write them, nor write a NN in this case).

Answer 1

RBM's do not do prediction tasks. They are generative models. You can use the transform method to get a hidden state transformation of the input, or your gan use the gibbs method to sample from the network.

You will then use randomized hill climbing algorithm to find good weights for a neural network.

No, this is not available in scikit-learn.

It sounds like your assignment might be meant for you to implement a simpler problem from scratch rather than use another library, as hill climbing isn't normally used for training a neural network. And they probably don't want you to do hill climbing for an RBM neural network.

You should probably consult your professor for more direction on what you really should be doing.