Header logo is ei

Bounded Rational Decision-Making in Feedforward Neural Networks


Conference Paper


Bounded rational decision-makers transform sensory input into motor output under limited computational resources. Mathematically, such decision-makers can be modeled as information-theoretic channels with limited transmission rate. Here, we apply this formalism for the first time to multilayer feedforward neural networks. We derive synaptic weight update rules for two scenarios, where either each neuron is considered as a bounded rational decision-maker or the network as a whole. In the update rules, bounded rationality translates into information-theoretically motivated types of regularization in weight space. In experiments on the MNIST benchmark classification task for handwritten digits, we show that such information-theoretic regularization successfully prevents overfitting across different architectures and attains results that are competitive with other recent techniques like dropout, dropconnect and Bayes by backprop, for both ordinary and convolutional neural networks.

Author(s): Leibfried, F and Braun, D
Book Title: Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence (UAI)
Pages: 407--416
Year: 2016
Month: June

Department(s): Empirical Inference
Bibtex Type: Conference Paper (conference)

Event Place: New York, NY, USA


  title = {Bounded Rational Decision-Making in Feedforward Neural Networks},
  author = {Leibfried, F and Braun, D},
  booktitle = {Proceedings of the 32nd Conference on Uncertainty in Artificial Intelligence (UAI)},
  pages = {407--416},
  month = jun,
  year = {2016},
  month_numeric = {6}