Ensemble X Your personal strataGEM to build Ensemble
summary
A concrete pipeline builds an ensemble called EnsembleX that trains three pretrained networks plus a custom CNN on OCT skin lesion and malaria sets uses data augmentation trains models individually then freezes base layers assigns nonlinear weights and fine tunes the top dense layers with the Adam optimizer over 250 epochs The ensemble reports about 89.5% accuracy with 1.47% misclassification and outputs confidence scores to reduce variance improve reliability and mitigate overfitting
AI-written summary · about
chapters (15)
- 0:00Welcome to EnsembleX
- 2:54Medical imaging advancements
- 4:18Problem identification
- 6:03Variance challenges
- 8:28EnsembleX workflow
- 10:13Model architectures
- 13:46Key contributions
- 14:07EnsembleX and the central limit theorem
- 16:33Reliability in EnsembleX
- 17:36Performance analysis
- 18:40Overfitting mitigation
- 20:04Medical use results
- 22:30Comparative study
- 24:18Future work
- 27:49Closing remarks
related talks
25mEnsemble-X: Your personal strataGEM to build Ensembled Deep Learning Models for Medical Imaging
20mMulti-algorithm Ensemble Learning at Scale: Software...
44mEnsemble Techniques for High Performance Machine Learning
32mEnsembles of GANs as a Data Augmentation Technique for Alzheimer research
22mA Scalable, Ensemble Approach for Building and Visualizing Deep Code-Sharing Networks
11mTree Ensemble Classifiers on heterogeneous platforms
31mIndividual Gaps in Prediction Generalization
40mPutting Deep Hybrid Learning to Work