A Dual Input-aware Factorization Machine for CTR Prediction
Authors: Wantong Lu, Yantao Yu, Yongzhe Chang, Zhen Wang, Chenhui Li, Bo Yuan
IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Comprehensive experiments on two real-world CTR prediction datasets show that the DIFM model can outperform several state-of-the-art models consistently. |
| Researcher Affiliation | Academia | 1Shenzhen International Graduate School, Tsinghua University 2Data 61 CSIRO, Sydney, Australia 3UBTECH Sydney AI Centre, The University of Sydney, Australia |
| Pseudocode | No | No pseudocode or clearly labeled algorithm blocks were found in the paper. |
| Open Source Code | No | The paper mentions 'We implement all models using Tensorflow4' and provides a general link to TensorFlow's website (https://www.tensorflow.org/), but does not explicitly state that the source code for their proposed DIFM model is publicly available or provide a link to it. |
| Open Datasets | Yes | Avazu2 dataset was published in the contest of Avazu Click-Through Rate Prediction in 2014. It contains click logs with 40 millions of data instances. ... We follow the data processing details of IFM [Yu et al., 2019]... Criteo3 dataset contains one month of ad click logs. ... 2http://www.kaggle.com/c/avazu-ctr-prediction 3http://labs.criteo.com/downloads/ download-terabyte-click-logs/ |
| Dataset Splits | No | For the Avazu dataset, the paper states '80% is for training and 20% is for testing'. For the Criteo dataset, it mentions '7 consecutive days of samples for training, and the next 1 day for evaluation'. A specific separate validation split is not explicitly mentioned with percentages or counts. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware used for running its experiments, such as GPU or CPU models, or memory specifications. |
| Software Dependencies | No | The paper states 'We implement all models using Tensorflow' and refers to the Adam optimizer, but does not provide specific version numbers for TensorFlow or any other software dependencies needed to replicate the experiment. |
| Experiment Setup | Yes | Parameter Settings. We implement all models using Tensorflow4. To enable a fair comparison, all models are learned by optimizing the log loss (Equation 14) using the Adam (Learning rate: 0.001) optimizer [Kingma and Ba, 2014]. For all methods, the embedding sizes for Criteo and Avazu are set to 20 and 40, respectively, which is the same as the parameter settings in IFM [Yu et al., 2019]. The batch size is set to 2000 for both datasets. The default setting for the number of neurons per layer is: (1) 256 for DNNs layers in Deep FM, x Deep FM, IFM and DIFM; (2) 200 for CIN layers in x Deep FM. For AFM, the attention factor is also set to 256 as the original paper recommended [Xiao et al., 2017]. ... we fixed the number of heads to 16 for the following experiments. ... we fixed the size of attention factor to 100 for Avazu and 80 for Criteo. ... relu is indeed more appropriate for neurons of the vector-wise part. |