ComparingRegressionwithDeep LearningandSVMBy: Peshal Pokhrel, Qudrat Ratul, Brett Sneed, and MaxwellVestrand
OutlineDefinitionsWhat doRegression,SVM, andDeep Learningshare?What isRegression?What isDeep Learning?What isSupport Vector Machine (SVM)?RecommendationsWhen isSVMa better first choice thanRegression?When isRegressiona better first choice thanSVM?When isDeep Learninga better first choice thanRegression?When isRegressiona better first choice thanDeep Learning?Conclusion
What do Regression, SVM, and Deep Learning share?●Each is a Supervised Learningmodel: develop function fromlabelled training data whichapproximately maps inputdata to output data.
What is Regression?w
Get answer to your question and much more
What is SVM?●Quadratic Programmingproblem with constraints●In typical real-world problems,classes will non-separable.●In this case, have a trade-offbetween maximizing marginR()and minimizing trainingerrorE(,).●Hinge loss:
What is Deep Learning?●Each vector of connectionweightswmaps to aconnection between layers●Deep learning architectureincluding activation functionsare encapsulated inf(,)●To be true “Deep” learning,need at least say 10 layers
When is SVM a better first choice thanLogistic Regression?