TaylorBoost: First and second-order boosting algorithms with explicit margin control

Saberian, Mohammad; Masnadi-Shirazi, Hamed; Vasconcelos, Nuno

doi:10.1109/cvpr.2011.5995605

Cited by 27 publications

(25 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With smaller trees, MCDB gives the best results on all datasets indicating that MCDB only requires very weak base classifiers even if there is no exact weak learner condition for MCDB. GD-MCBoost achieves the second best accuracy, and this algorithm also requires weaker base classifiers since it is able to boost any type of weak learners with non-zero directional derivatives [14]. We do not report its results on Poker525k dataset since its one iteration takes more than 12 hours to run by authors' matlab code.…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Semi-supervised Discriminative Structured Prediction

Wang¹

2013

View full text Add to dashboard Cite

Assembling a slave object into a fixture-free master object represents a critical challenge in flexible manufacturing. Existing deep reinforcement learning-based methods, while benefiting from visual or operational priors, often struggle with smallbatch precise assembly tasks due to their reliance on insufficient priors and high-costed model development. To address these limitations, this paper introduces a cognitive manipulation and learning approach that utilizes skill graphs to integrate learningbased object detection with fine manipulation models into a cohesive modular policy. This approach enables the detection of the master object from both global and local perspectives to accommodate positional uncertainties and variable backgrounds, and parametric residual policy to handle pose error and intricate contact dynamics effectively. Leveraging the skill graph, our method supports knowledge-informed learning of semisupervised learning for object detection and classroom-to-real reinforcement learning for fine manipulation. Simulation experiments on a gear-assembly task have demonstrated that the skillgraph-enabled coarse-operation planning and visual attention are essential for efficient learning and robust manipulation, showing substantial improvements of 13% in success rate and 15.4% in number of completion steps over competing methods. Real-world experiments further validate that our system is highly effective for robotic assembly in semi-structured environments.

show abstract

Section: Resultsmentioning

confidence: 99%

“…For comparison, we also report the results of AdaBoost.M1 [6], AdaBoost.MH [16], SAMME [30], and GD-MCBoost [14]. All these algorithms use multi-class base classifiers except AdaBoost.MH, which reduces the multi-class problem to a set of binary classification problems.…”

Section: Resultsmentioning

confidence: 99%

Semi-supervised Discriminative Structured Prediction

Wang¹

2013

View full text Add to dashboard Cite

show abstract

“…Using a Taylor series expansion of R(f k + g) around f k , the first and second order functional variations along the direction of weak learner g(x) ∈ G are [18] δR(f…”

Section: Boostingmentioning

confidence: 99%

“…The proposed boosting algorithm is based on the TaylorBoost framework [18]. Assume that after k iterations the predictor has m terms,…”

Section: Boosting Sum Of Products Sop-boostmentioning

confidence: 99%

“…A new framework for the design of boosting algorithms that learns such combinations is then introduced. The framework is simple and mathematically sound, relying on the statistical view of Boosting [6] and Taylor series approximations in functional space [18]. The resulting boosting algorithms grow a predictor by selecting among a pair of predefined operations, which could be sums and products, or "ands" and "ors", among others.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation