Train and Test Tightness of LP Relaxations in Structured Prediction
Authors: Ofer Meshi, Ben London, Adrian Weller, David Sontag
JMLR 2019 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section we present some numerical results to support our theoretical analysis. We run experiments for a multi-label classification task and an image segmentation task. For training we have implemented the block-coordinate Frank-Wolfe algorithm for structured SVM (Lacoste-Julien et al., 2013), using GLPK as the LP solver. We use a standard L2 regularizer, chosen via cross-validation. Fig. 4 shows relaxed and exact training iterations for the Yeast dataset (14 labels). We plot the relaxed and exact hinge terms (Eq. (6)), the exact and relaxed SSVM training objectives12 (Eq. (3) and Eq. (4), respectively), fraction of train and test instances having integral solutions, as well as test accuracy (measured by F1 score). |
| Researcher Affiliation | Collaboration | Ofer Meshi EMAIL Google Ben London EMAIL Amazon Adrian Weller EMAIL University of Cambridge and The Alan Turing Institute David Sontag EMAIL MIT CSAIL |
| Pseudocode | No | The paper describes algorithms like the block-coordinate Frank-Wolfe algorithm but does not present any structured pseudocode or algorithm blocks with numbered steps or a dedicated 'Algorithm' section. |
| Open Source Code | No | The paper does not contain any explicit statements about releasing source code, nor does it provide links to code repositories. |
| Open Datasets | Yes | For multi-label classification we adopt the experimental setting of Finley and Joachims (2008). In this setting labels are represented by binary variables, the model consists of singleton and pairwise factors forming a fully connected graph over the labels, and the task loss is the normalized Hamming distance. Fig. 4 shows relaxed and exact training iterations for the Yeast dataset (14 labels). ... We conduct experiments on a foreground-background segmentation problem using the Weizmann Horse dataset (Borenstein et al., 2004). |
| Dataset Splits | Yes | Finally, we conduct experiments on a foreground-background segmentation problem using the Weizmann Horse dataset (Borenstein et al., 2004). The data consists of 328 images, of which we use the first 50 for training and the rest for testing. |
| Hardware Specification | No | The paper mentions using GLPK as an LP solver but does not specify any hardware components like CPU, GPU models, or memory for the experimental setup. |
| Software Dependencies | No | For training we have implemented the block-coordinate Frank-Wolfe algorithm for structured SVM (Lacoste-Julien et al., 2013), using GLPK as the LP solver. We use a standard L2 regularizer, chosen via cross-validation. The paper mentions 'GLPK as the LP solver' but does not provide a specific version number for it or any other software dependency. |
| Experiment Setup | No | We use a standard L2 regularizer, chosen via cross-validation. For multi-label classification we adopt the experimental setting of Finley and Joachims (2008). The paper mentions the type of regularizer and refers to another paper for experimental settings, but does not explicitly provide specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or other detailed training configurations in the main text. |