Instinctual learning

As explained in Evolutionary Deep Learning: Genetic algorithms and neural networks

Instinctual learning is a mental model in which inherited or evolved responses guide behavior without conventional individual learning at the time of action.

Instinctual learning (IL) is an abstract model for using evolution to develop generalized machine-learning behavior. An instinct is a core learned behavior: it may emerge through evolution, or through reinforcement and repeated practice until a response becomes automatic. IL uses that transition as a design pattern for digital agents that reuse behavior across tasks. section 12.3 section 12.3.1

The motivating analogy is dual-process thought. A slower, conscious, trial-and-error process resembles system 2; with sufficient conditioning, its behavior can become a fast system-1 response. IL treats this shift from learned behavior to reusable instinct as a starting point rather than a complete theory of generalized intelligence. section 12.3.1

An IL-style implementation separates task-specific parts from reusable behavior:

• Represent candidate behaviors as functions, operators, or network-layer blocks that can be selected and evolved. section 12.0 section 12.6
• Keep input and output layers specific to each environment while sharing middle-layer instinct slots. section 12.6
• Evaluate an individual across multiple environments, then evolve layer choices toward a shared instinct pool. section 12.6
• Allow reused functions to be mixed and matched, while constraining where they may appear in the network. section 12.6

Shared instincts in a DQN agent

A pair of DQN networks can solve two environments with separate environment-facing layers and three shareable middle-layer slots. A candidate gene sequence selects layer blocks for each agent; the selected blocks are assembled with the Keras functional API, evaluated across environments, and evolved. The setup demonstrates generalization by seeking a shared base set of behaviors, not by making every layer identical. section 12.6 section 12.7

IL is exploratory and deliberately simplified. Breaking a DQN into instincts can demonstrate cross-task reuse, but generalized digital life and broadly generalized ML remain difficult goals rather than guaranteed outcomes. section 12.3.1 section 12.7

Figure 12.15 IL applied to DL

This diagram shows the central architectural move: two DQN agents retain environment-specific edges while drawing from shareable instinct slots in their middle layers. section 12.6

How Evolutionary Deep Learning: Genetic algorithms and neural networks covers it

How deep it goes

Explains the behavioral analogy, then implements shared DQN layer slots and evolutionary evaluation without claiming a complete general-intelligence theory.

Examples

Demonstrates conditioned dog behavior and a multi-environment DQN with evolved reusable middle-layer blocks.

What it assumes

Assumes familiarity with DQN and deep reinforcement learning while introducing IL through biological and conditioning analogies.

What you can do after

Build it from scratch by assembling environment-specific DQN edges around evolved shared layer slots.

How it connects

Connects evolutionary computation, reinforcement learning, deep reinforcement learning, DQN, genetic programming, and generalized machine learning.

liveBookmanning.com

Read Evolutionary Deep Learning: Genetic algorithms and neural networks on liveBook Get the book at manning.com

Related concepts

• deep reinforcement learninguses · Deep reinforcement learning is reinforcement learning that uses deep neural networks to represent policies, value functions, or other decision-making components.
• Functional APIuses · The Functional API is a Keras interface for constructing neural-network models as directed computation graphs by connecting layers and tensors.
• evolutionary computationuses · Evolutionary computation is a family of search and optimization methods that iteratively improve candidate solutions through variation and selection inspired by biological evolution.
• deep learninguses · Deep learning is a machine-learning approach that uses neural networks with multiple computational layers to learn representations from data.
• reinforcement learninguses · Reinforcement learning is a machine-learning paradigm in which an agent learns actions or a policy through interaction with an environment and evaluative feedback, such as rewards.

This explainer was generated by Manning's knowledge graph from the text of Evolutionary Deep Learning: Genetic algorithms and neural networks. The section links point to the parts of the book it draws on.