A Generative Model for Deep Convolutional Learning

TL;DR

This paper introduces a generative deep convolutional model with a novel probabilistic pooling mechanism, enabling effective multi-layer feature learning and achieving high classification accuracy on benchmark datasets.

Contribution

It presents a new probabilistic pooling operation within a deep generative framework for convolutional dictionary learning, enhancing feature learning and classification performance.

Findings

Effective multi-layer feature learning from images

High classification accuracy on MNIST and Caltech 101 datasets

Demonstrates the power of probabilistic pooling in deep models

Abstract

A generative model is developed for deep (multi-layered) convolutional dictionary learning. A novel probabilistic pooling operation is integrated into the deep model, yielding efficient bottom-up (pretraining) and top-down (refinement) probabilistic learning. Experimental results demonstrate powerful capabilities of the model to learn multi-layer features from images, and excellent classification results are obtained on the MNIST and Caltech 101 datasets.