class: center, middle, inverse, title-slide # Shiny adaboosting: an interactive dashboard to adaptive boosting algorithm ## IV International Seminar on Statistics with R ### ⚔ <br/> Mateus Maia <br/> <br/> <small> a joint work with Anderson Ara </small> <br/> <br/> <br/> Federal University of Bahia. ### 2019-05-23 --- # Statistical Learning Models <br/> <br/> .center[] --- class: inverse, center, middle # Ensemble Models --- # Ensemble Models .center[] --- # Ensemble Models <br/> <br/> .center[] --- # Boosting Models .center[] --- class: inverse, center, middle # Adaptative Boosting ## (AdaBoosting) --- # AdaBoosting Given `$$\mathbf{y}\in\{1,-1\}$$` <br/> <br/> `$$G(\mathbf{x})=sign \left(\sum_{m=1}^{M} \alpha_{m}g_{m}(\mathbf{x}) \right)$$` -- <br/> <br/> ##.center[The weighted wisdom of a crowd of experts!] --- # AdaBoosting <br/> <br/> .center[] --- # AdaBoosting <br/> <br/> .center[] --- # AdaBoosting Applications - Optical Character Recognition Task (Drucker, Schapire and Simard ,1993) -- - Text filtering and classification (Lee, et.al, 2011) -- - Speech recognition (Saon, 2012) -- - Object detection (Chen, 2011) -- - And so on... --- background-size: cover class: center, middle, inverse # ShinyAdaBoosting: They application --- background-image: url(shiny2.png) background-size: 100px background-position: 90% 8% #Shiny <br/> <br/> ###.center["Shiny is an open source R package that provides an elegant and powerful web framework for building web applications using R. (RStudio Team, 2019)"] --- class: inverse, center, middle # ShinyAdaBoosting: knowing the application .center[] --- # ShinyAdaBoosting: Setting the parameters -Database -Training Ratio -Number of Models -Type of Tree Model -Animated Model -Animate? --- class: inverse, center, middle # ShinyAdaBoosting: Vizualizing the Data .center[] --- # ShinyAdaBoosting: Final Model .center[] --- class: inverse, center, middle # ShinyAdaBoosting: All Models .center[] --- class: inverse, center, middle # ShinyAdaBoosting: Animated Mode .center[] --- # When Stumps aren't enough .center[] --- # Complete Trees AdaBoosting .center[] --- class: inverse, center, middle # All Models Complete Trees .center[] --- class: inverse, center, middle # Animate Complete Trees AdaBoost .center[] --- # More models? .center[] --- ## References -DRUCKER, Harris; SCHAPIRE, Robert; SIMARD, Patrice. __Boosting performance in neural networks.__ In:Advances in Pattern Recognition Systems using Neural Network Technologies. 1993. p. 61-75 -LEE, J. J., LEE, P. H., LEE, S. W., YUILLE, A., & KOCH, C. (2011, September). __Adaboost for text detection in natural scene.__ In 2011 International Conference on Document Analysis and Recognition (pp. 429-434). IEEE. -SAON, George; SOLTAU, Hagen. __Boosting systems for large vocabulary continuous speech recognition.__ Speech communication, v. 54, n. 2, p. 212-218, 2012. -CHEN, Shi et al. __Boosting part-sense multi-feature learners toward effective object detection.__ Computer Vision and Image Understanding, v. 115, n. 3, p. 364-374, 2011. --- ##Acknowledgments <br/> <br/> .center[] --- class: center, middle,inverse # Thank you! mateusmaia11@gmail.com @MateusMaiaM <br/> ### Explore a little more the ShinyAdaBosting app in mateusmaia.shinyapps.io/adaboosting/ .footnote[ [1]Slides created via the R package xaringan ]