Deep Lattice Networks and Partial Monotonic Functions
Authors: Seungil You, David Ding, Kevin Canini, Jan Pfeifer, Maya Gupta
NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on benchmark and real-world datasets show that six-layer monotonic deep lattice networks achieve state-of-the art performance for classification and regression with monotonicity guarantees. [...] Experiments on benchmark and real-world scenarios in Section 6. |
| Researcher Affiliation | Industry | Seungil You, David Ding, Kevin Canini, Jan Pfeifer, Maya R. Gupta Google Research 1600 Amphitheatre Parkway, Mountain View, CA 9043 {siyou,dwding,canini,janpf,mayagupta}@google.com |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our implementation is open sourced and can be found in https://github.com/tensorflow/lattice. |
| Open Datasets | Yes | We present results on the same benchmark dataset (Adult) with the same monotonic features as in Canini et al. [3], and for three problems from Google where the monotonicity constraints were specified by product groups. ... 6.2 Adult Benchmark Dataset (Classification) We compare accuracy on the benchmark Adult dataset [19]... [19] C. Blake and C. J. Merz. UCI repository of machine learning databases, 1998. |
| Dataset Splits | Yes | Table 2: Dataset Summary ... Adult Classify 90 (4) 26,065 6,496 16,281 User Intent Classify 49 (19) 241,325 60,412 176,792 Rater Score Regress 10 (10) 1,565,468 195,530 195,748 Usefulness Classify 9 (9) 62,220 7,764 7,919 ... We randomly split the usual train set [19] 80-20 and trained over the 80%, and validated over the 20%. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions 'Tensor Flow [15]' and 'Adam optimizer [16]' but does not provide specific version numbers for these or other software dependencies. |
| Experiment Setup | Yes | For classification problems, we used logistic loss, and for the regression, we used squared error. For each problem, we used a validation set to optimize the hyperparameters for each model architecture: the learning rate, the number of training steps, etc. For an ensemble of lattices, we tune the number of lattices, G, and number of inputs to each lattice, S. All calibrators for all models used a fixed number of 100 keypoints, and set [ 100, 100] as an input range. In all experiments, we use the six-layer DLN architecture: Calibrators Linear Embedding Calibrators Ensemble of Lattices Calibrators Linear Embedding, and validate the number of lattices in the ensemble G, number of inputs to each lattice, S, the Adam stepsize and number of loops. |