Coursework, Assignments and relevant codes created for the Advanced Machine Learning course at the University of Texas at Austin.
Includes -
Classical papers of techniques such as SGD with adaptive learning rates and momentum, conjugate gradient methods, point processes as well as newer methods and implementations such as counterfactuals for explainable AI.
Reviewed basic concepts such as regularization, bias-variance trade-offs, SGD, RMSProp etc. Explored newer ideas, eg. COPOD in outlier detection, DiCE for counterfactuals, Bayesian Inference, Calibrations in a multitude of regression and classification settings using MLPs, boosted decision trees etc.
Charter of techniques in explainable AI and their performance and stability in deep learning models. Learnt basis for techniques such as Integrated Gradients, SmoothGrad etc. in addition to the classics Lime and SHAP.
Details published on Medium article : https://medium.com/@shivarjunsarkar/explainable-ai-821365505974
