Research map / Learning theory and algorithms / Prediction and decision losses
RoBoSS: A Robust, Bounded, Sparse, and Smooth Loss Function for Supervised Learning
Mushir Akhtar, M. Tanveer, Mohd. Arshad
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use RoBoSS loss instead of pinball loss (best baseline) | SVM pin loss · 79 UCI and KEEL datasets (D1 category) | average accuracy (%): 84.26 -> 86.35 | Improved |
| use RoBoSS loss instead of wave loss (best baseline) | SVM wave loss · EEG datasets (32 tasks) | average accuracy (%): 74.73 -> 79.42 | Improved |
| use RoBoSS loss instead of LINEX loss (best baseline) | SVM LINEX loss · BreaKHis datasets (16 tasks) | average accuracy (%): 60.41 -> 63.25 | Improved |
| use RoBoSS loss instead of hinge loss | SVM hinge loss · 79 UCI and KEEL datasets (D1 category) | average training time (s): 0.1304 -> 0.0012 | Improved |
| use RoBoSS loss instead of pinball loss | SVM pin loss · 79 UCI and KEEL datasets (D1 category) | average rank: 2.96 -> 2.16 | Improved |
| use RoBoSS loss instead of pinball loss (Nemenyi test) | SVM pin loss · 79 UCI and KEEL datasets (D1 category) | statistical significance: rank difference 0.8, No | No clear effect |
| use RoBoSS loss instead of hinge loss (Nemenyi test) | SVM hinge loss · 79 UCI and KEEL datasets (D1 category) | statistical significance: rank difference 1.19, Yes | Improved |
| use RoBoSS loss instead of qtse loss | SVM qtse loss · EEG datasets (32 tasks) | average accuracy (%): 55.19 -> 79.42 | Improved |
| use RoBoSS loss instead of qtse loss in rank | SVM qtse loss · EEG datasets (32 tasks) | average rank: 6 -> 1.06 | Improved |
| use RoBoSS loss instead of hinge loss on BreaKHis | SVM hinge loss · BreaKHis datasets (16 tasks) | average rank: 4.22 -> 1.38 | Improved |
| use RoBoSS loss instead of qtse loss (BreaKHis time) | SVM qtse loss · BreaKHis datasets (16 tasks) | average training time (s): 0.0011 -> 0.0016 | Hurt |
| vary loss hyperparameters a and lambda (ablation) | RoBoSS-SVM · abalone9-18, echocardiogram, titanic, ecoli3 | accuracy | Mixed |
Closest papers by meaning
- Solving L1-regularized SVMs and Related Linear Programs: Revisiting the Effectiveness of Column and Constraint Generation (JMLR 2022 · similarity 0.87)
- Approximation Vector Machines for Large-scale Online Learning (JMLR 2017 · similarity 0.85)
- Supervised learning: no loss no cry (ICML 2020 · similarity 0.85)
- A Unified View on Multi-class Support Vector Classification (JMLR 2016 · similarity 0.85)
- Sparse SVM with Hard-Margin Loss: a Newton-Augmented Lagrangian Method in Reduced Dimensions (JMLR 2025 · similarity 0.85)
- EnsLoss: Stochastic Calibrated Loss Ensembles for Preventing Overfitting in Classification (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
