v0.6.1

Change Log

Major Features and Improvements

• Fix bug in CCPM model that incorrectly using linear_feature_columns as DNN input features
• Fix bug in DynamicGRU, a shape mismatch error will be caused in previous version when using DIEN with number of history_feature not equal to 2

v0.6.0

Change Log

Major Features and Improvements

• Now DeepCTR is compatible with tensorflow 1.14 and 2.0.

v0.5.2

Change Log

Major Features and Improvements

Refactor Linear Layer

v0.5.1

Change Log

Major Features and Improvements

• Add FiBiNET Chinese Introducton FiBiNET：结合特征重要性和双线性特征交互进行CTR预估
  

v0.5.0

Change Log

Major Features and Improvements

• Refactor inputs module,now you can input different features to linear part and dnn part of the models.
• Dense value inputs are concated with the inputs of dnn instead of embed into dense vectors.
• previous versions of the code require modification to work properly,please refer new examples to update your codes.

v0.4.1

Change Log

Major Features and Improvements

• Add DSIN(Deep Session Interest Network)
  paper code
• Refactor layers

v0.4.0

Change Log

Major Features and Improvements

• Previous versions of the code require a little modification to work properly
• Support feature hashing on the fly in training and inference process example
• Add python2.7 support
• Refactor methods in input_embedding.py
• Refactor layers and models

API changes

• deepctr.utils.SingleFeatand deepctr.utisl.VarlenFeat add new arguments to support feature hashing
• deepctr.core.MLP to deepctr.core.DNN
• arguments of all models and layers changes

v0.3.4	v0.4.0
l2_reg_deep	l2_reg_dnn
activation	dnn_activation
keep_prob	dnn_dropout
hidden_size	dnn_hidden_units
use_bn	dnn_use_bn
final_activation	task

v0.3.4

Change Log

Major Features and Improvements

• Add FGCNN(Feature Generation by Convolutional Neural Network)CCPM&FGCNN:使用CNN进行特征生成的CTR预测模型
• Add FGCNNLayer in deepctr.layers.interaction doc
• Now we can use deepctr under tensorflow 1.13, but there may be some compatibility issues.

v0.3.3

Change Log

Major Features and Improvements

• Add CCPM(Convolutional Click Prediction Model)
• Add KMaxPooling in deepctr.layers.sequence doc

v0.3.2

Change Log

Major Features and Improvements

• Add DIEN
• Add NFFM