Why best performance MSE does not align with final MSE?

2 次查看（过去 30 天）

显示更早的评论

Sunita 2023-12-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2063747-why-best-performance-mse-does-not-align-with-final-mse

评论： pathakunta 2024-1-26

ANN model outputs

The model stopped at epoch 22 with an MSE of 0.0242
The best performance was observed at epoch 16 with an MSE of 0.031
However, the final MSE between actual and predicted values is 0.0308

Shouldn't the third data align with the best performance (second) value ?

5 个评论
显示 3更早的评论隐藏 3更早的评论

Mahi 2023-12-24

MSE at epoch 22 is better than the epoch 16. There is always a difference between training and testing error and accuracy values . Your final value of MSE is higher than your training mse .which I think is good.otherwise it may mean that your model is over fitted

pathakunta 2024-1-26

The model stopped at epoch 22 with an MSE of 0.0242

请先登录，再进行评论。

请先登录，再回答此问题。

回答（1 个）

Ganesh 2023-12-24

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2063747-why-best-performance-mse-does-not-align-with-final-mse#answer_1377107

移动：Matt J 2023-12-25

MSE you achieve at epoch 22 is for that specific minibatch used at epoch 22. That would not mean that you achieve the same MSE for all of the dataset. It's quite possible that you achieve a lower or a higher MSE on the test set.