Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
On Robustness and Regularization of Structural Support Vector Machines
Authors: Mohamad Ali Torkamani, Daniel Lowd
ICML 2014 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our experimental results show that our method outperforms the nonrobust structural SVMs on real world data when the test data distribution has drifted from the training data distribution. |
| Researcher Affiliation | Academia | Mohamad Ali Torkamani EMAIL Daniel Lowd EMAIL Computer and Information Science Department, University of Oregon |
| Pseudocode | No | The paper does not contain any structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | The expanded political blogs dataset and our robust SVM implementation can be downloaded from the following URL: http://ix.cs.uoregon.edu/ lowd/ robustsvmstruct. |
| Open Datasets | Yes | We introduce a new dataset based on the political blogs dataset collected by Adamic and Glance (2005). |
| Dataset Splits | Yes | We partitioned the blogs into three separate sub-networks and used three-way cross-validation, training on one subnetwork, using the next as a validation set for tuning parameters, and evaluating on the third. |
| Hardware Specification | No | The paper mentions using the Gurobi optimization engine but does not specify any hardware details like CPU, GPU, or memory. |
| Software Dependencies | Yes | We learned parameters using a cutting plane method, implemented using the Gurobi optimization engine 5.60 (2014) for running all integer and quadratic programs. |
| Experiment Setup | Yes | Standard structural SVMs have one parameter C that needs to be tuned. The robust method has an additional regularization parameter C = 1/Be = 1/Bw which scales the strength of the robust regularization. We chose these parameters from the semi-logarithmic set {0, .001, .002, .005, .1, . . . , 10, 20, 50}. We learned parameters using a cutting plane method, implemented using the Gurobi optimization engine 5.60 (2014) for running all integer and quadratic programs. We ran for 50 iterations and selected the weights from the iteration with the best performance on the tuning set. |