What is an Environment?

The environment is a simulation which encompasses the dynamics of the market and calculatesd the effects of the agent’s rebalancing in the portfolio value. It must be a Gymnasium object implementing the following methods:

• reset: Changes the environment’s state to the initial state.

• step: Responsible for running a simulation step given an input action defined by the reinforcement learning agent.

• render: Returns informations that can be used to render the environment.

Currently, there is one environment implemented in the library called PortfolioOptimizationEnv, and it will be detailed in the next page.