Function approximator
A function approximator is a computational model, typically parameterized, that represents or estimates a mapping from inputs to outputs.
Applied Reinforcement Learning
Function approximation estimates a value function or policy with a parameterized mathematical function instead of storing a separate value for every state or state-action pair.
- How it is used
- Replaces tabular value storage when continuous, high-dimensional, or combinatorial state spaces make enumeration impractical.
- How deep it goes
- Explains parameter adjustment with temporal-difference error and traces deep-network learning through layers, nonlinear activations, and backpropagation.
- Examples
- Uses radiotherapy dose planning with tumor volume, biological indicators, accumulated dose, and patient-specific factors as state variables.
- What you can do after
- Evaluate/understand when linear models with engineered features or deep networks for raw inputs are appropriate.
Read how Applied Reinforcement Learning explains function approximator →
Math and Architectures of Deep Learning
The task of representing a target input-output mapping with a model, used here to justify nonlinear models for complex real-world processes.
- How it is used
- Frames machine learning as approximating unknown classification or estimation functions from training data.
- How deep it goes
- Connects surrogate functions, weights, nonlinear activations, perceptron geometry, and approximation capacity, while deferring practical parameter-count limits.
- Examples
- Uses stock-price estimation, logical gates, XOR, sigmoid curves, and perceptron decision-boundary diagrams.
- What you can do after
- Build it from scratch, formulate weighted models and perceptrons that approximate simple mappings.
Read how Math and Architectures of Deep Learning explains function approximator →
Evolutionary Deep Learning: Genetic algorithms and neural networks
A function approximator is a computational model that estimates the outputs of a target function without necessarily reproducing its exact form.
- How it is used
- Frames machine-learning prediction and evolutionary search as alternative ways to approximate or construct functions.
- How deep it goes
- Defines the role, compares methods, and walks through a runnable Geppy implementation without developing mathematical theory.
- Examples
- Uses continuous polynomials, discontinuous and complex functions, and a Geppy target of 6*x1 + 22.
- What you can do after
- Build it from scratch by configuring Geppy to evolve, evaluate, simplify, and visualize an executable function.
What to know first
- reward signalA reward signal is a value or feedback indicating the desirability, benefit, or cost associated with an action, outcome, or behavior.
- actionAn action is a choice or operation selected or performed by an actor that can affect a system, environment, or process.
- convolutional neural networkA convolutional neural network is a neural-network architecture that processes structured inputs with convolutional operations, typically using learned local filters to capture patterns.
- stateState is information describing the condition or configuration of a system or computation at a particular time that may influence its behavior or subsequent transitions.
Related concepts
- deep neural networkimplemented by · A deep neural network is a neural-network model with multiple layers of computational units that learn representations from data.
- tabular methodalternative to · A reinforcement-learning method that stores value estimates or policy information explicitly in lookup tables indexed by states or state-action pairs.
- deep learningaddressed by · Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
- linear function approximationincludes · A function-approximation technique that estimates values as weighted sums of state or state-action features.
- machine learningequivalent to · Machine learning is a field of computational methods in which systems learn patterns or relationships from data to make predictions, decisions, or other outputs.
- DQNimplemented by · A deep Q-network (DQN) is a reinforcement-learning method that uses a neural network to approximate the state-action value function in Q-learning.
Questions about function approximator
What is function approximator?
A function approximator is a computational model, typically parameterized, that represents or estimates a mapping from inputs to outputs.
Which Manning books explain function approximator?
Applied Reinforcement Learning, Math and Architectures of Deep Learning and Evolutionary Deep Learning: Genetic algorithms and neural networks.
What should I know before learning function approximator?
Manning books that teach function approximator usually introduce reward signal, action, convolutional neural network and state first.
What is the difference between function approximator and tabular method?
A function approximator is a computational model, typically parameterized, that represents or estimates a mapping from inputs to outputs. A reinforcement-learning method that stores value estimates or policy information explicitly in lookup tables indexed by states or state-action pairs.
What kinds of function approximator are there?
Kinds of function approximator include linear function approximation.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.


