Abstract | ||
---|---|---|
Training a Neural Network (NN) with lots of parameters or intricate architectures creates undesired phenomena that complicate the optimization process. To address this issue we propose a first modular approach to NN design, wherein the NN is decomposed into a control module and several functional modules, implementing primitive operations. We illustrate the modular concept by comparing performances between a monolithic and a modular NN on a list sorting problem and show the benefits in terms of training speed, training stability and maintainability. We also discuss some questions that arise in modular NNs. |
Year | Venue | DocType |
---|---|---|
2019 | AAAI Spring Symposium - Combining Machine Learning with Knowledge Engineering | Journal |
Volume | Citations | PageRank |
abs/1902.09240 | 0 | 0.34 |
References | Authors | |
22 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
David Castillo-Bolado | 1 | 0 | 0.34 |
Cayetano Guerra-Artal | 2 | 0 | 0.34 |
Mario Hernández-tejera | 3 | 28 | 6.32 |