Deep Reinforcement Learning for Cryptocurrency Market Making

Summary

The paper outlines a deep reinforcement learning framework for cryptocurrency market making, with agents learning to manage inventory while interacting with a simulated trading environment. The environment represents market observations using limit order book data and order flow arrival statistics. Policy gradient algorithms serve as the agents, and a feed-forward neural network approximates their policies or value functions.

The experiment compares two reward functions and evaluates agent-reward combinations using daily and average trade returns. The authors present the framework as a way to address the stochastic inventory control problem faced by market makers. The supplied description does not identify the specific algorithms, reward definitions, dataset, or numerical results, so it does not establish how performance transfers to live markets or other market conditions.

Key ideas

• The framework applies deep reinforcement learning to cryptocurrency market making.
• Agents observe limit order book information and order flow arrival statistics.
• Policy gradient methods interact with the market environment, using a feed-forward neural network.
• Two reward functions are compared using daily and average trade returns.
• The stated application is stochastic inventory control for market makers, though the supplied description omits implementation and transfer details.

Tags

# Deep Reinforcement Learning in Cryptocurrency Market Making


# Deep Reinforcement Learning in Cryptocurrency Market Making









This paper sets forth a framework for deep reinforcement learning as applied to market making (DRLMM) for cryptocurrencies. Two advanced policy gradient-based algorithms were selected as agents to interact with an environment that represents the observation space through limit order book data, and order flow arrival statistics. Within the experiment, a forward-feed neural network is used as the function approximator and two reward functions are compared. The performance of each combination of agent and reward function is evaluated by daily and average trade returns. Using this DRLMM framework, this paper demonstrates the effectiveness of deep reinforcement learning in solving stochastic inventory control challenges market makers face.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.