DeepFM: a factorization-machine based neural network for CTR prediction
Citations
1,317 citations
Cites background or methods from "DeepFM: a factorization-machine bas..."
...tation vector for the instance. MLP. Given the concatenated dense representation vector, fully connected layers are used to learn the combination of features automatically. Recently developed methods [4, 5, 10] focus on designing structures of MLP for better information extraction. Loss. The objective function used in base model is the negative log-likelihood function defined as: L = − 1 N Õ (x,y)∈S (ylogp(...
[...]
...cally extracts nonlinear relations among features and equals to the BaseModel. Wide&Deep needs expertise feature engineering on the input of the "wide" module. We follow the practice in [10] to take cross-product of user behaviors and candidates aswideinputs.Forexample,inMovieLensdataset,itrefersto the cross-product of user rated movies and candidate movies. •PNN[5]. PNN can be viewed as...
[...]
...ation function with complex MLP network, which enhances the model capability greatly. PNN[5] tries to capture high-order feature interactions by involving a product layer after embedding layer. DeepFM[10] imposes a factorization machines as "wide" module in Wide&Deep [4] with no need of feature engineering. Overall, these methods follow a similar model structure with combination of embed...
[...]
1,070 citations
560 citations
Cites methods from "DeepFM: a factorization-machine bas..."
...DeepFM [47] is an end-to-end model which seamlessly integrates factorization machine and MLP....
[...]
...MLP [2, 13, 20, 27, 38, 47, 53, 54, 66, 92, 95, 157, 166, 185], [12, 39, 93, 112, 134, 154, 182, 183] Autoencoder [34, 88, 89, 114, 116, 125, 136, 137, 140, 159, 177, 187, 207], [4, 10, 32, 94, 150, 151, 158, 170, 171, 188, 196, 208, 209]...
[...]
550 citations
550 citations
Cites background or methods from "DeepFM: a factorization-machine bas..."
...Representative models include FNN [44], PNN [30], DeepCross [36], NFM [11], DCN [39], Wide&Deep [4], and DeepFM [8]....
[...]
...The Wide&Deep [4] and DeepFM [8] models overcome this problem by introducing hybrid architectures, which contain a shallow component and a deep component with the purpose of learning both memorization and generalization....
[...]
...PNN [30] and DeepFM [8] modify the above architecture slightly....
[...]
...We re-use the symbols in [8], where red edges represent weight-1 connections (no parameters) and gray edges represent normal connections (network parameters)....
[...]
...Therefore, multi-field categorical form is widely used by related works [8, 30, 36, 39, 44]....
[...]
References
123,388 citations
33,597 citations
2,469 citations
2,460 citations
"DeepFM: a factorization-machine bas..." refers background or methods in this paper
...The FM component is a factorization machine, which is proposed in [Rendle, 2010] to learn feature interactions for recommendation....
[...]
...I R ] 1 3 M ar 2 01 7 (FM) [Rendle, 2010] model pairwise feature interactions as inner product of latent vectors between features and show very promising results....
[...]
...Such a method is hard to generalize to model high-order feature interactions or those never or rarely appear in the training data [Rendle, 2010]....
[...]
...(FM) [Rendle, 2010] model pairwise feature interactions as inner product of latent vectors between features and show very promising results....
[...]
1,960 citations
"DeepFM: a factorization-machine bas..." refers background or methods in this paper
...Several deep learning models are proposed in recommendation tasks other than CTR prediction (e.g., [Covington et al., 2016; Salakhutdinov et al., 2007; van den Oord et al., 2013; Wu et al., 2016; Zheng et al., 2016; Wu et al., 2017; Zheng et al., 2017])....
[...]
...[Salakhutdinov et al., 2007; Sedhain et al., 2015; Wang et al., 2015] propose to improve Collaborative Filtering via deep learning....
[...]