kgkolias
MSc Data Science @ City, University of London BSc Applied Informatics @ University of Macedonia, Greece
London, UK
Pinned Repositories
adversarially-motivated-intrinsic-goals-variation
qlearning-dddqn-lunar-lander-v2
Q-Learning Algorithm and Dueling Double Deep Q Network Applied in Reinforcement Learning Tasks Project descriptionProject Conducted for the Module Deep Learning : Optimization, developed with Python and Pytorch The project consists of two parts: 1. Application of the Q-Learning Algorithm in a relatively simple task, 2. Application of the Dueling Double Deep Q Network in the LunarLander-v2 environment provided by OpenAI. Q-Learning • Tried different policies; Boltzmann, Epsilon Greedy (Epsilon Decay and Constant) and Random • Hyperparameter Tuning and Optimization (Alpha and Gamma Parameters) • Created a Grid Environment Dueling Double Deep Q Network • Hyperparameter Tuning • Neural Network Architecture Optimization • E-greedy-decay policy implementation Tools and Technologies: Python, Pytorch, Pandas, Numpy, Matplotlib, OpenAI Gym
airbnbLondon-propertyPricing-data-analysis-and-predictions
Coursework for the module 'Principles of Data Science' for pricing of Airbnb properties in London.
fmnist-cnn-mlp-image-classification
A Comparative Study through Convolutional Neural Networks and Multilayer Perceptrons in Image Classification Comparison of Convolutional Neural Networks and Multilayer Perceptrons applied in a multi-class image classification problem, in the F-MNIST dataset. • The F-MNIST dataset, consists of 70,000 images; 60,000 used for Training and 10,000 for Test purposes • Implemented on Matlab. Project Details: Neural Network Architecture Optimization, Hyperparameter Tuning, Regularization, Grid Search, Image Generation using Python, Data Preprocessing, Data Analysis
naiveBayes-randonForest-wineClassification
teachDeepRL-variation
Thesis on proc generation of environments using ppo as a student on MiniGrid environments
kgkolias's Repositories
kgkolias/adversarially-motivated-intrinsic-goals-variation
kgkolias/teachDeepRL-variation
Thesis on proc generation of environments using ppo as a student on MiniGrid environments
kgkolias/qlearning-dddqn-lunar-lander-v2
Q-Learning Algorithm and Dueling Double Deep Q Network Applied in Reinforcement Learning Tasks Project descriptionProject Conducted for the Module Deep Learning : Optimization, developed with Python and Pytorch The project consists of two parts: 1. Application of the Q-Learning Algorithm in a relatively simple task, 2. Application of the Dueling Double Deep Q Network in the LunarLander-v2 environment provided by OpenAI. Q-Learning • Tried different policies; Boltzmann, Epsilon Greedy (Epsilon Decay and Constant) and Random • Hyperparameter Tuning and Optimization (Alpha and Gamma Parameters) • Created a Grid Environment Dueling Double Deep Q Network • Hyperparameter Tuning • Neural Network Architecture Optimization • E-greedy-decay policy implementation Tools and Technologies: Python, Pytorch, Pandas, Numpy, Matplotlib, OpenAI Gym
kgkolias/fmnist-cnn-mlp-image-classification
A Comparative Study through Convolutional Neural Networks and Multilayer Perceptrons in Image Classification Comparison of Convolutional Neural Networks and Multilayer Perceptrons applied in a multi-class image classification problem, in the F-MNIST dataset. • The F-MNIST dataset, consists of 70,000 images; 60,000 used for Training and 10,000 for Test purposes • Implemented on Matlab. Project Details: Neural Network Architecture Optimization, Hyperparameter Tuning, Regularization, Grid Search, Image Generation using Python, Data Preprocessing, Data Analysis
kgkolias/airbnbLondon-propertyPricing-data-analysis-and-predictions
Coursework for the module 'Principles of Data Science' for pricing of Airbnb properties in London.
kgkolias/naiveBayes-randonForest-wineClassification