Hybrid approach

As explained in Deep Learning with R, Third Edition

A hybrid system is a computational system that combines learned continuous components with discrete algorithmic components.

A hybrid system combines learned continuous components with discrete algorithmic components. Learning supplies perception and intuition; programs supply reasoning, search, memory, planning, or symbolic manipulation. Combining both can outperform either an end-to-end neural network or software without machine-learning elements. section 19.4.5 section 20.3

Hybrid designs commonly divide responsibilities between components:

• Deep-learning modules process data and provide pattern recognition, perception, intuition, or learned value and policy estimates. section 19.4.5 section 20.3
• Discrete programs handle structured operations such as search, reasoning, planning, memory, control flow, variables, recursion, and data structures. section 19.4.5 section 20.3
• The interface between them lets learned components update or guide an explicit model, while algorithmic components organize the resulting decisions. section 19.4.5

AlphaGo and autonomous driving

AlphaGo combines handcrafted Monte Carlo tree search with learned value and policy networks. A self-driving system can maintain a human-designed 3D world model while deep-learning perception modules continually update it from the vehicle’s surroundings. These divisions combine learned recognition with explicit algorithmic structure. section 19.4.5

Figure 19.15 A learned program relying on both geometric primitives (pattern recognition, intuition) and algorithmic primitives (reasoning, search, memory)

Use the figure to compare geometric primitives for pattern recognition and intuition with algorithmic primitives for reasoning, search, and memory. section 19.4.5

How Deep Learning with R, Third Edition covers it

How deep it goes

Explains component roles and system examples, but stops short of implementation details, training procedures, or formal architecture specifications.

Examples

Uses AlphaGo’s tree search with policy and value networks, plus autonomous driving’s 3D world model and learned perception.

What you can do after

Evaluate/understand how learned and algorithmic components divide responsibilities in high-performance AI systems.

How it connects

Connects deep learning with program synthesis, symbolic reasoning, program search, modular recombination, and recurrent neural networks.

liveBookmanning.com

Read Deep Learning with R, Third Edition on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Building Quantum Software with PythonA hybrid algorithm combines classical and quantum computation in a single procedure, typically using classical processing to guide repeated quantum computations.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionA model that combines deep-learning components with algorithmic or symbolic components.
• 

  liveBookmanning.com
  Bootstrapping Microservices, Second Edition: With Docker, Kubernetes, GitHub Actions, and TerraformThe hybrid approach is an application architecture that retains a monolith for most functionality while surrounding it with microservices where their flexibility or scalability is beneficial.
• 

  liveBookmanning.com
  Effective Conversational AIA hybrid conversational approach combines classification, predefined or task-oriented dialogue flows, and generated responses in one conversational system.

Compare how 5 books explain hybrid approach →

Related concepts

• algorithmic moduleuses · An algorithmic module is a reusable computational component that performs discrete, algorithmic, or symbolic operations such as reasoning, search, or abstraction.
• Grover adaptive searchincludes · Grover adaptive search is a hybrid quantum-classical search technique that varies the number of Grover-operator iterations according to a schedule when the number of desired outcomes is unknown.
• JSONuses · JSON is a standardized, language-independent text format for representing structured data using objects, arrays, strings, numbers, Boolean values, and null.
• fine-tuninguses · Fine-tuning is adapting a pretrained machine-learning model by training it further on targeted data or objectives for a task, domain, or behavior.

This explainer was generated by Manning's knowledge graph from the text of Deep Learning with R, Third Edition. The section links point to the parts of the book it draws on.