SARSAAgent<T>

Models & Types · AiDotNet.ReinforcementLearning.Agents.SARSA

SARSA (State-Action-Reward-State-Action) agent using tabular methods.

Inheritance: ReinforcementLearningAgentBase<T> → SARSAAgent<T>

Implements: IGradientComputable<T, Vector<T>, Vector<T>>

For Beginners

SARSA is like Q-Learning’s more cautious cousin. While Q-Learning learns the optimal policy assuming perfect future actions, SARSA learns based on what you actually do (including exploratory mistakes).

Key differences from Q-Learning:

  • On-Policy: Learns from actions it actually takes
  • More Conservative: Safer in risky environments (cliff walking)
  • Exploration Aware: Updates reflect exploration strategy
  • Convergence: Converges to optimal policy only if exploration decreases

Update rule: Q(s,a) ← Q(s,a) + α[r + γQ(s’,a’) - Q(s,a)] (Uses actual next action a’, not max)

Perfect for: Environments where safety matters, risky state transitions Famous for: Rummery & Niranjan 1994, on-policy TD control

How It Works

SARSA is an on-policy TD control algorithm that learns Q-values based on the action actually taken by the current policy, not the optimal action.

Example

// Create an on-policy SARSA agent for safe exploration
var options = new SARSAOptions<double> { LearningRate = 0.1, StateSize = 4, ActionSize = 2 };
var agent = new SARSAAgent<double>(options);
            
// Select an action for the current state
var state = new Vector<double>(new double[] { 0.5, -0.3, 1.0, 0.2 });
var action = agent.SelectAction(state);

Constructors

Constructor Summary
SARSAAgent()
SARSAAgent(SARSAOptions<T> options)

Properties

Property Summary
int FeatureCount { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
T Train()
Vector<T> ComputeGradients(Vector<T> input, Vector<T> target, ILossFunction<T> lossFunction)
Vector<T> SelectAction(Vector<T> state, bool training)
byte[] Serialize() Auto-generated common model serialization surface.
void ApplyGradients(Vector<T> gradients, T learningRate)
void Deserialize(byte[] data) Auto-generated common model deserialization surface.
void LoadModel(string filepath)
void ResetEpisode()
void SaveModel(string filepath)
void StoreExperience(Vector<T> state, Vector<T> action, T reward, Vector<T> nextState, bool done)