Skip to main content
. 2022 Oct 4;62(20):4852–4862. doi: 10.1021/acs.jcim.2c00715

Figure 4.

Figure 4

Siamese-MFBERT network architecture for both training strategies: classification (left) and augmented latent representations (right). The two MFBERT network weights are shared in both cases, providing the Siamese architecture.