omBot | Phân loại tin giả | VT Hùng, TK Nghĩa
Lựa chọn mô hình |
Cấu trúc:
Model: "BERT"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_ids (InputLayer) [(None, 256)] 0 []
attention_mask (InputLayer) [(None, 256)] 0 []
tf_roberta_model (TFRobertaMod TFBaseModelOutputWi 134998272 ['input_ids[0][0]',
el) thPoolingAndCrossAt 'attention_mask[0][0]']
tentions(last_hidde
n_state=(None, 256,
768),
pooler_output=(Non
e, 768),
past_key_values=No
ne, hidden_states=N
one, attentions=Non
e, cross_attentions
=None)
flatten (Flatten) (None, 196608) 0 ['tf_roberta_model[0][0]']
batch_normalization (BatchNorm (None, 196608) 786432 ['flatten[0][0]']
alization)
dense (Dense) (None, 64) 12582976 ['batch_normalization[0][0]']
dropout_37 (Dropout) (None, 64) 0 ['dense[0][0]']
outputs (Dense) (None, 2) 130 ['dropout_37[0][0]']
==================================================================================================
Total params: 148,367,810
Trainable params: 12,976,322
Non-trainable params: 135,391,488
__________________________________________________________________________________________________
N = 421 | Dự đoán là tin thật | Dự đoán là tin giả |
---|---|---|
Tin thật (204) | 193 | 10 |
Tin giả (217) | 13 | 205 |
N = 421 {'loss': 2.30495023727417, 'accuracy': 0.9360189437866211}
Model: "BiLSTM"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 512, 128) 1280000
bidirectional (Bidirectiona (None, 256) 263168
l)
dense (Dense) (None, 128) 32896
dense_1 (Dense) (None, 64) 8256
dense_2 (Dense) (None, 1) 65
=================================================================
Total params: 1,584,385
Trainable params: 1,584,385
Non-trainable params: 0
_________________________________________________________________
N = 421 | Dự đoán là tin thật | Dự đoán là tin giả |
---|---|---|
Tin thật (204) | 185 | 21 |
Tin giả (217) | 17 | 198 |
N = 421 {'loss': 0.26830601692199707, 'accuracy': 0.9099525809288025}
Model: "LSTM"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 512, 128) 1280000
lstm (LSTM) (None, 128) 131584
dense (Dense) (None, 128) 16512
dense_1 (Dense) (None, 64) 8256
dense_2 (Dense) (None, 1) 65
=================================================================
Total params: 1,436,417
Trainable params: 1,436,417
Non-trainable params: 0
_________________________________________________________________
N = 421 | Dự đoán là tin thật | Dự đoán là tin giả |
---|---|---|
Tin thật (204) | 168 | 43 |
Tin giả (217) | 35 | 175 |
N = 421 {'loss': 0.44977718591690063, 'accuracy': 0.8151658773422241}
Model: "GRU"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 512, 128) 1280000
gru (GRU) (None, 128) 99072
dense (Dense) (None, 128) 16512
dense_1 (Dense) (None, 64) 8256
dense_2 (Dense) (None, 1) 65
=================================================================
Total params: 1,403,905
Trainable params: 1,403,905
Non-trainable params: 0
_________________________________________________________________
N = 421 | Dự đoán là tin thật | Dự đoán là tin giả |
---|---|---|
Tin thật (204) | 154 | 52 |
Tin giả (217) | 44 | 171 |
N = 421 {'loss': 0.4920154809951782, 'accuracy': 0.7725118398666382}
Model: "CNN"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 512, 128) 1280000
conv1d (Conv1D) (None, 508, 128) 82048
flatten (Flatten) (None, 65024) 0
dense (Dense) (None, 128) 8323200
dense_1 (Dense) (None, 64) 8256
dense_2 (Dense) (None, 1) 65
=================================================================
Total params: 9,693,569
Trainable params: 9,693,569
Non-trainable params: 0
_________________________________________________________________
N = 421 | Dự đoán là tin thật | Dự đoán là tin giả |
---|---|---|
Tin thật (204) | 137 | 67 |
Tin giả (217) | 56 | 161 |
N = 421 {'loss': 0.5959132313728333, 'accuracy': 0.7085307836532593}