Efficient Learning of Convolution Weights as Gaussian Mixture Model Posteriors
This work addresses the challenge of unsupervised learning in convolutional neural networks for image modeling, though it appears incremental as it builds on existing Gaussian mixture and EM frameworks.
The paper tackled the problem of learning convolution weights efficiently without supervision by showing that a convolution layer's feature map corresponds to the unnormalized log posterior of a Gaussian mixture model, and it proposed an EM algorithm for learning that is guaranteed to converge.
In this paper, we showed that the feature map of a convolution layer is equivalent to the unnormalized log posterior of a special kind of Gaussian mixture for image modeling. Then we expanded the model to drive diverse features and proposed a corresponding EM algorithm to learn the model. Learning convolution weights using this approach is efficient, guaranteed to converge, and does not need supervised information. Code is available at: https://github.com/LifanLiang/CALM.