Applied Sciences, Vol. 14, Pages 8020: Efficient Implementation of Multilayer Perceptrons: Reducing Execution Time and Memory Consumption
Applied Sciences doi: 10.3390/app14178020
Authors: Francisco Cedron Sara Alvarez-Gonzalez Ana Ribas-Rodriguez Santiago Rodriguez-Yañez Ana Belen Porto-Pazos
A technique is presented that reduces the required memory of neural networks through improving weight storage. In contrast to traditional methods, which have an exponential memory overhead with the increase in network size, the proposed method stores only the number of connections between neurons. The proposed method is evaluated on feedforward networks and demonstrates memory saving capabilities of up to almost 80% while also being more efficient, especially with larger architectures.