The boundary of neural network trainability is fractal.
Jascha Sohl-DicksteinPublished in: CoRR (2024)
Keyphrases
- neural network
- back propagation
- fractal dimension
- pattern recognition
- artificial neural networks
- neural network model
- recurrent neural networks
- image compression
- object boundaries
- fuzzy neural network
- backpropagation neural network
- genetic algorithm
- image gradient
- fuzzy artmap
- prediction model
- feed forward
- neural network is trained
- image processing
- fractal analysis
- hybrid intelligent
- boundary extraction
- image reconstruction from projections
- learning vector quantization
- feedforward neural networks
- multi layer perceptron
- medial axis
- multi layer
- neural nets
- associative memory
- self organizing maps
- edge detection