Pinned Repositories
DogBreedApp2
Using transfer learning, this app takes the pre-trained convolutional neural net Xception, restructures it to only predict 133 different dog breeds, and trains the neural net on a little over 8,500 dog images. However, since the neural net is only trained to recognize dog image features, it is automatically capable of determining which dog breed a human image most closely resembles.
House-Price-Optimization
Used Decision Tree/Support Vector Regression to optimize house prices.
Improved-Q-Learning-Multi-Environment
Open A.I. Research: Developed Q-Learning algorithm with Computer Vision to learn two different video games with infinite state spaces, while optimizing for learning speed. Algorithm combined the use of a Convolutional Neural Network with Priority Sweeping to lower training time from 20 hours to 5 hours. Games used were Flappy Bird and Monster Kong.
Predicting-Likely-Donors
Used Support Vector Machine, Decision Tree, and Naive Bayes to predict likely donors for charity.
06kahao's Repositories
06kahao/Improved-Q-Learning-Multi-Environment
Open A.I. Research: Developed Q-Learning algorithm with Computer Vision to learn two different video games with infinite state spaces, while optimizing for learning speed. Algorithm combined the use of a Convolutional Neural Network with Priority Sweeping to lower training time from 20 hours to 5 hours. Games used were Flappy Bird and Monster Kong.
06kahao/DogBreedApp2
Using transfer learning, this app takes the pre-trained convolutional neural net Xception, restructures it to only predict 133 different dog breeds, and trains the neural net on a little over 8,500 dog images. However, since the neural net is only trained to recognize dog image features, it is automatically capable of determining which dog breed a human image most closely resembles.
06kahao/House-Price-Optimization
Used Decision Tree/Support Vector Regression to optimize house prices.
06kahao/Predicting-Likely-Donors
Used Support Vector Machine, Decision Tree, and Naive Bayes to predict likely donors for charity.