C-Index Optimization: Is Meta-Learning The Answer?
To robustify fast adaptation, this paper optimizes meta learning pipelines from a distributionally robust perspective and meta trains models with the measure of expected tail risk. Uses the chain rule in the opposite direction from backprop, accumulating derivatives from start to finish. Blackbox bayesian hyperparameter optimization. Operator ∈ {relu, leaky relu, tanh} {conv3x3, ∈ separable conv3x3, max pool,. } b = operation in layer k (conv3x3,. •deep learning works very well, but requires large datasets •in many cases, we only have a small amount of data available (e. g. , some specific computer.
