Skip to content

Latest commit

 

History

History
45 lines (23 loc) · 929 Bytes

README.md

File metadata and controls

45 lines (23 loc) · 929 Bytes

Imdb_Emotion_Classification

Imdb Dataset Emotion Classification

Just for practice...

Be careful... There might be overfitting.

-Accuracy in 20 epochs = 0.8639 . . .

-Accuracy in 4 epochs = 0.8840 -picked

Accuracy of the system changes by number of hidden layers,

-with one hidden layer (Fully Connected / relu) = 0.8837

-with two hidden layers (Fully Connected / relu) = 0.8840

-with three hidden layers (Fully Connected / relu) = 0.8874 -picked

Accuracy changes by number of units in layers

-with 16 unit on each hidden layer = 0.8874 -picked

-with 32 unit on each hidden layer = 0.8690

-with 64 unit on each hidden layer = 0.8354

Deferences between loss functions

-Accuracy when binary_crossentropy used = 0.8874 -picked

-Accuracy when mse used = 0.8839

Deferences between activation functions

-Accuracy when relu used = 0.8874 -picked

-Accuracy when tanh used = 0.8667