Hello author, thank you for your contribution. I trained a model using 100 whole slide images (WSI) from 70 patients' pathology, and I noticed that the c-index of the test dataset dropped from 0.6 to 0.26. Is this because I have too little data and the model is overfitting? I did not follow the process of using level=1, 2 but used level=0, 1 instead. Is this overfitting, or could there be other reasons?
[training] training epoch 29, avg. batch loss: 1.96642664, loss: 0.90510726, c_index: 0.72959
[training] validation epoch 29, loss: 0.92027694, c_index: 0.90909
[training] test epoch 29, loss: 0.91841877, c_index: 0.42105
superli6
commented
4 days ago
Hello author, thank you for your contribution. I trained a model using 100 whole slide images (WSI) from 70 patients' pathology, and I noticed that the c-index of the test dataset dropped from 0.6 to 0.26. Is this because I have too little data and the model is overfitting? I did not follow the process of using level=1, 2 but used level=0, 1 instead. Is this overfitting, or could there be other reasons? [training] training epoch 29, avg. batch loss: 1.96642664, loss: 0.90510726, c_index: 0.72959 [training] validation epoch 29, loss: 0.92027694, c_index: 0.90909 [training] test epoch 29, loss: 0.91841877, c_index: 0.42105
[training] training epoch 69, avg. batch loss: 1.29691067, loss: 0.31053680, c_index: 0.89467 [training] validation epoch 69, loss: 0.80728710, c_index: 0.90909 [training] test epoch 69, loss: 1.07747674, c_index: 0.26316
[training] training epoch 92, avg. batch loss: 1.15009958, loss: 0.15829915, c_index: 0.95858 [training] validation epoch 92, loss: 0.75507718, c_index: 0.90909 [training] test epoch 92, loss: 0.85490167, c_index: 0.26316