1

CNNs often surpass RBMs in most computer vision tasks in terms of classification accuracy. However, has there been any analysis comparing the following factors:

  • Hyper-parameter sensitivity
  • Amount of training data required
  • Computational power/time required

Note that I do understand that RBMs and CNNs are two completely different algorithms, as covered in this question. However, they can and often are, both used for classification. Even though in classification, the RBM is mostly just used as a feature extractor for an SVM.

Community
  • 1
Seanny123
  • 625
  • 4
  • 20
  • 1
    Note that I usually ask these very application-heavy questions on cs.stackexchange.com, but I thought I would try this site this time. – Seanny123 Dec 08 '16 at 11:23

1 Answers1

2

If RBM reference to a Restricted Boltzmann Machine than there is no reason to compare RBM and CNN, because they are completely different algorithms. CNN is supervised algorithm and RBM is unsupervised. They are trying to achieve different goals in different ways.

itdxer
  • 7,104
  • 1
  • 18
  • 28