Glossary

Healthcare in 3D Terms and Phrases

  • Action (Machine Reinforcement Learning) — The decision that could be made by the Clinician (Agent) at the Inflection Point.
  • Agent (Machine Reinforcement Learning) — In HCn3D represent the clinician (and patient) taking Action at the Inflection Point.
  • Discount factor (Machine Reinforcement Learning) — Provides a method for weighting or biasing the value of immediate vs. deferred Rewards. In HCn3D it is used to influence Outcomes for acute vs chronic care. For instance an acute condition would necessarily require a bias towards more immediate Actions.
  • Environment (Machine Reinforcement Learning) — The world through which the Agent moves. In HCn3D it is the Healthcare Network and all data associated with the Patient Record, both intrinsic and extrinsic. Computationally it is where the Agent’s Actions are computed to formulate the feedback, the Reward and next iteration.
  • Policy (Machine Reinforcement Learning) — Is the strategy employed by the Agent as it moves through the Environment to determine its next Action.
  • Q-value (Action value) (Machine Reinforcement Learning) — Is the prediction of the Q Function which formulates the highest probable Reward from the sequences of State/Action pairs, essentially a consideration of all Action Spectrums from the current cube through the last cube achieving the desired Trajectory. It is HCn3D best “guess” for the desired future outcomes at the current Inflection Point. Together with Policy the application of Counterfactual analysis.
  • Reinforcement Learning — In HCn3D is an iterative form of “Semi-Supervised Learning” whereby feedback is counterfactually analyzed for the next iteration. The process evaluates its actions based on outcomes compared with known patient journeys. Its aim is to learn sequences of actions that will achieve the goal of predicting the patient’s possible futures (Future Value Outcomes)
Page 1 of 2 1 2