I’m getting training accuracy and testing accuracy much greater than 1, with training accuracy around 64 and test accuracy around 32. I’m using gloss.SoftmaxCrossEntropyLoss and linear regression model with one hidden layer from chapter 3. My batch size is 256. I split the features and labels when doing part 3. Want to ask why this is happening?
Related Topics
Topic | Replies | Views | Activity | |
---|---|---|---|---|
How to normalize the softmax and how the accuracy works?
|
6 | 3446 | September 13, 2018 | |
HW5 Q1.4: Accuracies | 0 | 336 | February 25, 2019 | |
Getting ~60% accuracy with new categorization | 0 | 456 | March 1, 2019 | |
HW5 Q2 and Q3 | 7 | 655 | February 26, 2019 | |
Optimal hyperparameters for training resnet34_v1 on ImageNet? | 4 | 2117 | July 19, 2018 |