Fig. 2: Classification architecture of the prognostic prediction model. | Communications Medicine

Fig. 2: Classification architecture of the prognostic prediction model.

From: A convolutional attention model for predicting response to chemo-immunotherapy from ultrasound elastography in mouse tumor models

Fig. 2

The prediction G of the last layer of the network is combined with the L1 and L2 estimators in the second and third layers, respectively. Attention weights a1 and a2 are created by applying a softmax function to the compatibility scores c1 and c2. Attention mechanism outputs G1 and G2 are finally concatenated to create the final prediction of the model.

Back to article page