Experimental Setups Sample Clauses

Experimental Setups. After an extensive hyper-parameter search, the statistics of hyper-parameters used in the proposed GCN models are shown in Table 5.2. Additionally, we set the maximum sequence length to 128 for the transformer encoder. Different seed values are used for the three runs and the average accuracy on development set and test set are calculated. 64 1e-3 1000 300 0.6 Table 5.2: Statistics of hyper-parameters. BS: batch size; LR: learning rate; E: number of training epochs; HC: size of hidden channel of GCN; DR: dropout rate.