Learning Monotonic Probabilities with a Generative Cost Model
Authors: Yongxiang Tang, Yanhua Cheng, Xiaocheng Liu, Chenchen Jiao, Yanxiang Zeng, Ning Luo, Pengjia Yuan, Xialong Liu, Peng Jiang
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We further validate our approach with a numerical simulation of quantile regression and conduct multiple experiments on public datasets, showing that our method significantly outperforms existing monotonic modeling techniques. |
| Researcher Affiliation | Industry | 1Kuaishou, Beijing, China. Correspondence to: Yongxiang Tang <EMAIL>, Yanhua Cheng <EMAIL>. |
| Pseudocode | No | The paper describes the methodology using mathematical formulations and textual descriptions rather than structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The code for our experiments can be found at https://github.com/tyxaaron/GCM. |
| Open Datasets | Yes | To further evaluate the monotonic problem with a multivariate revenue variable r, we conduct experiments on six publicly available datasets: the Adult dataset (Becker & Kohavi, 1996), the COMPAS dataset (Larson et al., 2016), the Diabetes dataset (Teboul), the Blog Feedback dataset (Buza, 2014), the Loan Defaulter dataset and the Auto MPG dataset (Quinlan, 1993). |
| Dataset Splits | Yes | The datasets are divided into training and testing sets in a 4 : 1 ratio. |
| Hardware Specification | No | The paper does not provide specific details regarding the hardware used for running the experiments, such as GPU or CPU models. |
| Software Dependencies | No | The paper mentions that network parameters are optimized using the stochastic gradient descent algorithm and refers to ADAM (Kingma & Ba, 2015) as an optimizer, but does not provide specific version numbers for software libraries or dependencies like Python, PyTorch, or TensorFlow. |
| Experiment Setup | Yes | Table 5 (Hyperparameters of the experiments) lists specific hyperparameters for each dataset, including hidden dimension, sample number, latent dimension, max epoch, optimizer, batch size, and learning rate. For example, for the Adult dataset, it specifies a batch size of 256, learning rate of 0.001, and a max epoch of 40. |