
Abstract: This target of this workshop is twofold. On one hand, it is familiarizing attendees with mechanics of reinforcement learning (RL) applied to financial environments. On the other side, it aims to uncover key differences between popular RL applications (as playing video games) and financial ones, ignoring which inevitably will lead to losses of time and capital. With such insights and code boilerplates, attendees will be able to avoid harsh mistakes and implement environment-driven strategies faster.
Session Outline
This workshop will be split into 3 parts:
- First, we will quickly review fundamentals of RL and will apply it to a historical cryptocuurencies market dataset (turned into an environment) as it was a simple game
- Then, we will dive deeper and realize why such an approach won't ever become a viable trading strategy and how advances in financial machine learning can fix that
- At the end, we will implement such fixes and will see how it makes our previously developed framework more reliable in the real-wold conditions.
The workshop will end with the discussion and a roadmap for studying and applying ML and RL in finance.
Background Knowledge
- Intermediate knowledge of machine learning and underlying mathematics is preferred to follow the tutorial
- Scientific computing skills are necessary to follow the tutorial steps (Numpy, Scipy, Pandas, TF)
- Experience with reinforcement learning is welcomed but is not necessary
- Experience with financial markers is welcomed but is not necessary
Bio: Alex Honchar is a tech entrepreneur and educator. Currently, he is co-founder and ML director at Neurons Lab - a consulting firm specializing in healthcare, finance, and IoT. Also, he writes a popular blog on Medium about machine learning applications and leadership. Previously he worked as an independent consultant with SMBs and startups on rapid go-to-market ML solutions and taught machine learning courses at the University of Verona and Ukrainian Catholic University.