Yiorgos Christakis

The project involved the comparison of strengths and weaknesses of breadth first search, uniform cost search, and A* search in uni/bi/tri directional implementations. The algorithms were each used to determine possible and then optimal routes through Romanian cities in a contrived example, and then again to determine similar routes for navigation from address to address in the city of Atlanta, Georgia.

Created a crypto currency investment tracker for personal use using python and web interactions with the coinmarketcap API. The investment tracker would execute on request to perform the following tasks:
Log a new trade
Report current holdings
Report current profit
Report lifetime profit
Report individual performance of trades
Report overall performance of trades

Applied Hidden Markov Models to the AI problem of decoding noisy morse code messages. The project involved implementing the HMM in python and using it to generate the most likely message given the received morse code.

Designed and executed a test for changes in auditory perception due to high dB sound using an EEG cap to analyze the n100 event related potential.

Designed an AI agent to play the game "Isolation" using adversarial search algorithms. Simple minimax was implemented, as well as minimax with alpha beta pruning and iterative deepening. The designed agent's minimum functional requirement was to defeat a random player greater than 90% of the time.

Used particle filter tracking in python to track various objects of interest through videos.

Created a simple portfolio website to learn HTML, CSS, and javascript.

Built a market simulator in python that accepts dataframes representing suggested trades, and simulates their outcomes over historical data. The market simulator was later used to train and test a library of algorithms including simple decision tree learners, random forests, regression learners, and qlearners.

Used Qlearning to create a trading policy meant to outperform benchmark stock performance, measured as simple buy and hold performance of a single stock. Using three technical indicators, rate of change(ROC), weighted moving average(WMA), and Bollinger Bands(BB%) a qlearner was developed to maximize daily returns by training on historical data.

Programmed a controller for a test robot to navigate itself to a given destination automatically using MATLAB and the Robotarium package.

Using second language acquisition to develop a learning framework, an android application called the russian reader was developed to aid new intermediate learners climb the initial barrier to becoming comfortable with reading in the Russian language.

Designed and programmed a pitch detector and simulated vocoder. Both LPC and CELP Vocoders were implemented to compare effectiveness at signal reproduction and efficiency.