STSGCN实验结果

[zhoujiajie-z]

在使用PEMSD4进行STSGCN实验时，得出的结果很差，MAE等数值非常大，选择的是average模式，请问我是指令出现了错误吗 输入的指令为：python run_model.py --task traffic_state_pred --model STSGCN --dataset PEMSD4 --gpu_id 1

[Kazeya27]

参数是没有问题的，可以提供一些别的信息吗

[zhoujiajie-z]

1 42.449322 inf 4294.089844 65.529305 42.301456 0.490499 4237.257324 65.094215 0.827990 0.828200 2 42.794460 inf 4498.738770 67.072639 42.641479 0.475649 4435.363770 66.598526 0.819767 0.821981 3 43.057289 inf 4475.805176 66.901459 42.884789 0.476247 4415.407715 66.448532 0.820681 0.821155 4 43.248966 inf 4421.492188 66.494301 43.082691 0.461589 4353.379395 65.980141 0.822838 0.823864 5 43.665028 inf 4567.639160 67.584312 43.448811 0.496027 4500.414551 67.085129 0.817009 0.818473 6 44.065445 inf 4570.409180 67.604805 43.863903 0.491554 4497.889160 67.066307 0.816895 0.817069 7 44.611374 inf 4736.529297 68.822449 44.406525 0.498664 4670.591797 68.341728 0.810279 0.810280 8 45.086155 inf 4810.409180 69.357117 44.878792 0.497615 4742.878906 68.868561 0.807308 0.807343 9 45.453972 inf 4937.063965 70.264244 45.284546 0.502918 4877.696289 69.840508 0.802221 0.802240 10 48.280487 inf 5525.532227 74.333923 48.080097 0.586792 5467.903809 73.945274 0.778710 0.778805 11 47.880833 inf 5450.558594 73.827896 47.633385 0.527917 5374.683105 73.312233 0.781763 0.781763 12 47.427937 inf 5210.931641 72.186783 47.217602 0.527962 5138.762207 71.685158 0.791426 0.791883

这个是STSGCN的结果，这些数值感觉有点不对

[sci-forthcoming]

这数值太有问题了，我的也是。不知道那个方面出现了问题

[Mayor-W]

python run_model.py --task traffic_state_pred --model STSGCN --dataset PEMSD3 我在训练过程中第28个epoch时损失函数突然增大后不再收敛： ... 2024-09-27 14:44:32,203 - INFO - Epoch [22/100] train_loss: 64.0900, val_loss: 62.9793, lr: 0.001000, 67.08s 2024-09-27 14:45:29,986 - INFO - epoch complete! 2024-09-27 14:45:29,987 - INFO - evaluating now! 2024-09-27 14:45:38,929 - INFO - Epoch [23/100] train_loss: 63.5692, val_loss: 61.1685, lr: 0.001000, 66.72s 2024-09-27 14:45:39,053 - INFO - Saved model at 23 2024-09-27 14:45:39,053 - INFO - Val loss decrease from 61.7892 to 61.1685, saving to ./libcity/cache/83156/model_cache/STSGCN_PEMSD3_epoch23.tar 2024-09-27 14:46:37,832 - INFO - epoch complete! 2024-09-27 14:46:37,833 - INFO - evaluating now! 2024-09-27 14:46:46,831 - INFO - Epoch [24/100] train_loss: 62.4819, val_loss: 74.6267, lr: 0.001000, 67.78s 2024-09-27 14:47:44,480 - INFO - epoch complete! 2024-09-27 14:47:44,480 - INFO - evaluating now! 2024-09-27 14:47:53,449 - INFO - Epoch [25/100] train_loss: 62.8727, val_loss: 62.0785, lr: 0.001000, 66.62s 2024-09-27 14:48:52,907 - INFO - epoch complete! 2024-09-27 14:48:52,907 - INFO - evaluating now! 2024-09-27 14:49:01,532 - INFO - Epoch [26/100] train_loss: 62.7499, val_loss: 59.2922, lr: 0.001000, 68.08s 2024-09-27 14:49:01,582 - INFO - Saved model at 26 2024-09-27 14:49:01,582 - INFO - Val loss decrease from 61.1685 to 59.2922, saving to ./libcity/cache/83156/model_cache/STSGCN_PEMSD3_epoch26.tar 2024-09-27 14:49:59,793 - INFO - epoch complete! 2024-09-27 14:49:59,793 - INFO - evaluating now! 2024-09-27 14:50:08,488 - INFO - Epoch [27/100] train_loss: 61.9447, val_loss: 60.7979, lr: 0.001000, 66.91s 2024-09-27 14:51:06,323 - INFO - epoch complete! 2024-09-27 14:51:06,323 - INFO - evaluating now! 2024-09-27 14:51:15,368 - INFO - Epoch [28/100] train_loss: 1265.1051, val_loss: 116.2739, lr: 0.001000, 66.88s 2024-09-27 14:52:13,837 - INFO - epoch complete! 2024-09-27 14:52:13,837 - INFO - evaluating now! 2024-09-27 14:52:22,198 - INFO - Epoch [29/100] train_loss: 117.3606, val_loss: 116.2638, lr: 0.001000, 66.83s 2024-09-27 14:53:19,853 - INFO - epoch complete! 2024-09-27 14:53:19,853 - INFO - evaluating now! 2024-09-27 14:53:28,792 - INFO - Epoch [30/100] train_loss: 117.3605, val_loss: 116.2641, lr: 0.001000, 66.59s 2024-09-27 14:54:26,957 - INFO - epoch complete! 2024-09-27 14:54:26,958 - INFO - evaluating now! 2024-09-27 14:54:36,073 - INFO - Epoch [31/100] train_loss: 117.3547, val_loss: 116.2833, lr: 0.001000, 67.28s 2024-09-27 14:55:33,981 - INFO - epoch complete! 2024-09-27 14:55:33,981 - INFO - evaluating now! ...

最后得到的结果也很差： MAE MAPE MSE RMSE masked_MAE masked_MAPE masked_MSE masked_RMSE R2 EVAR 1 59.446613 inf 7742.617188 87.992142 59.619953 0.842482 7776.875000 88.186592 0.628819 0.650195 2 59.099960 inf 8333.264648 91.286720 59.241497 0.916476 8370.964844 91.492973 0.600523 0.629567 3 59.331520 inf 8147.683594 90.264519 59.466206 0.936365 8182.406738 90.456657 0.609436 0.633534 4 59.251080 inf 8064.135254 89.800529 59.351322 1.019226 8096.790527 89.982170 0.613456 0.631617 5 59.599438 inf 7813.224609 88.392448 59.735882 0.951667 7846.640625 88.581268 0.625499 0.645311 6 60.805088 inf 8165.230469 90.361664 60.870480 1.157570 8195.560547 90.529335 0.608653 0.625571 7 60.399315 inf 7978.080078 89.320099 60.513092 1.061806 8011.001465 89.504196 0.617647 0.636387 8 60.575157 inf 8264.590820 90.909798 60.667603 1.092198 8298.078125 91.093788 0.603935 0.624145 9 61.126614 inf 8525.250000 92.332283 61.248249 0.987657 8562.075195 92.531479 0.591462 0.616512 10 61.163643 inf 8631.472656 92.905716 61.258816 1.069559 8667.003906 93.096748 0.586389 0.611168 11 61.859146 inf 8519.511719 92.301201 62.032639 0.882416 8559.208984 92.515991 0.591775 0.623537 12 62.262905 inf 8572.358398 92.587029 62.440941 0.887979 8613.141602 92.807014 0.589272 0.617788