-
Notifications
You must be signed in to change notification settings - Fork 494
Neural Factorization Machines
潜心 edited this page Sep 13, 2020
·
2 revisions
Neural Factorization Machines for Sparse Predictive Analytics
创新:将Embedding集合通过池化,转化为一个向量。Bi-Interaction层不需要额外的模型学习参数,更重要的是它在一个线性的时间内完成计算
原文笔记:https://mp.weixin.qq.com/s/1en7EyP3C2TP3-d4Ha0rSQ
采用Criteo数据集进行测试。数据集的处理见utils
文件,主要分为:
- 考虑到Criteo文件过大,因此可以通过
read_part
和sample_sum
读取部分数据进行测试; - 对缺失数据进行填充;
- 对密集数据
I1-I13
进行归一化处理,对稀疏数据C1-C26
进行重新编码LabelEncoder
; - 整理得到
feature_columns
; - 切分数据集,最后返回
feature_columns, (train_X, train_y), (test_X, test_y)
;
class NFM(keras.Model):
def __init__(self, feature_columns, hidden_units, dnn_dropout=0., activation='relu', bn_use=True, embed_reg=1e-4):
"""
NFM architecture
:param feature_columns: A list. dense_feature_columns + sparse_feature_columns
:param hidden_units: A list. Neural network hidden units.
:param activation: A string. Activation function of dnn.
:param dnn_dropout: A scalar. Dropout of dnn.
:param bn_use: A Boolean. Use BatchNormalization or not.
:param embed_reg: A scalar. The regularizer of embedding.
"""
- file:Criteo文件;
- read_part:是否读取部分数据,
True
; - sample_num:读取部分时,样本数量,
5000000
; - test_size:测试集比例,
0.2
; - embed_dim:Embedding维度,
8
; - dnn_dropout:Dropout,
0.5
; - hidden_unit:DNN的隐藏单元,
[256, 128, 64]
; - learning_rate:学习率,
0.001
; - batch_size:
4096
; - epoch:
10
;
采用Criteo数据集中前500w
条数据,最终测试集的结果为:AUC:0.776491