Since the Resnet-18 feature extractor uses a ReLU activation function, the feature representation values are all non-negative, so the inputs to tukey's ladder of powers transformation are all valid. As expected, the performance of both methods drops a bit when the pre-training is not done on the similar classes. Still FeCAM outperforms NCM by about 10% on the final accuracy. In Algorithm 1, we present the pseudo code for using FeCAM classifier.Algorithm 1 FeCAM Require: Training data (D

Euclidean metric is suboptimal and that feature distributions are heterogeneous.

However,itsextension tosolvingstochastic min-max problems withstrong convexity and strong concavity still remains open, and itisstill unclear whether a fast rate ofO(1/T) for the duality gapis achievable for stochastic min-max optimization under strong convexity and strong concavity.

The visualization results revealed the semantic similarity between categories. Furthermore, Figure 2 shows the projected sample featureg at different iterations of training. Therefore, the probability density off not only depends on its orientation but also its strength. In this way,{π,µ} were updated via the following E-stepandtheM-step. This section provides more discussions on the quantification of knowledge points.