Automatic Incisive-teeth Selection And Extraction
(Computer Vision) Given a training set of panoramic dental X-Rays, a program was designed and trained to locate and extract the 4 upper incisor teeth automatically.
(Recursive Least Squares Support Vector Machines) For predicting future values on a Time-series, I used Least Squares Support Vector Machines as a powerful non-linear autoregressive model.
Feedforward Artificial Neural Network, complete Python implementation using only Numpy and Scipy
(Artificial Neural Networks) This is my take in programing a Feedforward Artificial Neural Network with tanh activation functions and cross-entropy loss function WITHOUT using any ML packages. In this implementation I also programmed:
1) L2 regularization, 2) Drop-out regularization, 3) the same network using TensorFlow.
Images Classification, using Convolutional Neural Networks and ensambles of learners
(Deep Learning) I built a Conv-NN for classifying images of dogs by breed (available here). I used a One-vs-All approach to make every learner (Conv-NN), in an ensamble of learners, to correctly identify one breed from the rest. The complete ensamble then sees every image and gives a probability of belonging to the class. The image is assigned to the class with highest probability.