Tversky as a Loss Function for Highly Unbalanced Image Segmentation using 3D Fully Convolutional Deep Networks.
Seyed Raein HashemiSeyed Sadegh Mohseni SalehiDeniz ErdogmusSanjay P. PrabhuSimon K. WarfieldAli GholipourPublished in: CoRR (2018)
Keyphrases
- loss function
- image segmentation
- pairwise
- deep learning
- learning to rank
- logistic regression
- support vector
- markov random field
- regularization term
- empirical risk
- risk minimization
- boosting framework
- stochastic gradient descent
- hinge loss
- spectral clustering
- reproducing kernel hilbert space
- segmentation algorithm
- sparse coding
- expectation maximization
- objective function
- image processing