Applying Machine Learning for Stock Trading with Elixir

Sam McDavid presents a talk on leveraging Elixir for stock trading using machine learning techniques. He starts by dismissing common misconceptions about guaranteed profits and underscores the inherent risks in stock trading. The talk is structured around understanding traders' decision-making using technical analysis and translating this strategy into a usable machine learning policy. McDavid delves into using LSTM (Long Short-Term Memory) for time series data but finds it inadequate. He then shifts focus to reinforcement learning, particularly Q-learning, explaining how agents observe the environment, take actions, and receive rewards. He outlines the exploration versus exploitation dilemma and introduces Epsilon Decay as a solution. McDavid explains how to implement these concepts in Elixir using GenServers and concludes that while the models face challenges like model degradation, continuous refinement can yield more profitable outcomes. He emphasizes that these models require substantial capital and are more suited for hedge funds than individual investors.