Fig. 5
Performance evaluation by selecting different numbers of GO terms for GOAE and GONN model. a, b Average ARI and NMI results between GOAEv model and GOAE model. GOAEv represent GOAE model without selection of GO terms. c, d Average ARI and NMI results between GONNv model and GONN model. GONNv represent GONN model without selection of GO terms. GOAE models in (a) and (b) select different hyperparameters. Before GO terms selection, it selects epoch number=300, learning rate=1e-4. After selection, it selects epoch number=100, learning rate=1e-3. All GONN models in (c) and (d) all select epoch number=200 and learning rate=0.2. All these hyperparameters selection are achieve their best across a lots of hyperparameter groups