Corpus ID: 34916084

On Characterizing the Capacity of Neural Networks using Algebraic Topology

@article{Guss2018OnCT,
  title={On Characterizing the Capacity of Neural Networks using Algebraic Topology},
  author={William H. Guss and R. Salakhutdinov},
  journal={ArXiv},
  year={2018},
  volume={abs/1802.04443}
}
The learnability of different neural architectures can be characterized directly by computable measures of data complexity. In this paper, we reframe the problem of architecture selection as understanding how data determines the most expressive and generalizable architectures suited to that data, beyond inductive bias. After suggesting algebraic topology as a measure for data complexity, we show that the power of a network to express the topological complexity of a dataset in its decision… Expand
45 Citations
Neural Persistence: A Complexity Measure for Deep Neural Networks Using Algebraic Topology
  • 30
  • PDF
Understanding Deep Learning: Expected Spanning Dimension and Controlling the Flexibility of Neural Networks
Topology of deep neural networks
  • 9
  • PDF
Topologically Densified Distributions
  • 2
  • PDF
THE EXPRESSIVE POWER OF NEURAL NETWORKS
  • 3
  • PDF
Characterizing the Shape of Activation Space in Deep Neural Networks
  • 5
Path Homologies of Deep Feedforward Networks
  • 6
  • PDF
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 35 REFERENCES
On the Expressive Power of Deep Neural Networks
  • 368
  • PDF
Exponential expressivity in deep neural networks through transient chaos
  • 324
  • PDF
Toward Deeper Understanding of Neural Networks: The Power of Initialization and a Dual View on Expressivity
  • 189
  • PDF
On the number of response regions of deep feed forward networks with piece-wise linear activations
  • 125
  • PDF
Neural Architecture Search with Reinforcement Learning
  • 2,384
  • PDF
The Power of Depth for Feedforward Neural Networks
  • 429
  • PDF
PathNet: Evolution Channels Gradient Descent in Super Neural Networks
  • 366
  • PDF
Approximation by superpositions of a sigmoidal function
  • G. Cybenko
  • Mathematics, Computer Science
  • Math. Control. Signals Syst.
  • 1989
  • 4,161
  • PDF
...
1
2
3
4
...