AI Roadmap
ONE of FIVE features of the Mozaic Roadmap.
Continue to build upon the current self-taught RL (Reinforcement Learning) agent designed to navigate the complexities of yield farming in the DeFi ecosystem.
INCORPORATING DOMAIN KNOWLEDGE
Use domain experts to define possible black swan scenarios, even if they haven't occurred in the past. Integrating this knowledge into the environment simulation to expose the AI agent to these scenarios during training.
ANOMOLY DETECTION
Train a separate model (or system) for anomaly detection. When this system detects an anomalous state(s) in the market, which could be indicative of trend change, it can trigger the RL agent to switch to a more conservative strategy.
HUMAN-IN-THE-LOOP (HITL):
Especially for high-stakes areas like financial markets, having a mechanism where unusual decisions by the AI agent (that might indicate it's moving into unknown territory) are flagged for human review before procedure can be invaluable.
Last modified 16d ago