Performance analysis of associative memory of Restricted Boltzmann Machine

Performance of associative memory of Restricted Boltzmann Machine (RBM) presented in this letter. Some simulations are given based on the single layer and multi-layer RBM. In single layer test, RBM has better performance of associative memory with the 64 neurons in hidden layer, and in multi-layer test, RBM has strong robust memory performance when the layer is 3. The simulation results show that RBM of multi-layer can memory and restore image from the memory model even if the re-constructed image is added much noise, and in the meantime, it can be balanced between the running time and performance of memory through adjusting the number of layer and neurons amount of hidden layer.