evolvable 1.0.0 → 1.2.0

data/README.md

@@ -1,286 +1,359 @@
-# Evolvable
+# Evolvable 🦎
 
-A framework for building evolutionary behaviors in Ruby.
+[![Gem Version](https://badge.fury.io/rb/evolvable.svg)](https://badge.fury.io/rb/evolvable) [![Maintainability](https://api.codeclimate.com/v1/badges/7faf84a6d467718b33c0/maintainability)](https://codeclimate.com/github/mattruzicka/evolvable/maintainability)
 
-[Evolutionary algorithms](https://en.wikipedia.org/wiki/Evolutionary_algorithm) build upon ideas such as natural selection, crossover, and mutation to construct relatively simple solutions to complex problems. This gem has been used to implement evolutionary behaviors for [visual, textual, and auditory experiences](https://projectpag.es/evolvable) as well as a variety of AI agents.
+An evolutionary framework for writing programs that use operations such as selection, crossover, and mutation. Explore ideas generatively in any domain, discover novel solutions to complex problems, and build intuitions about intelligence, complexity, and the natural world.
 
-With a straightforward and extensible API, Evolvable aims to make building simple as well as complex evolutionary algorithms fun and relatively easy.
+Subscribe to the [Evolvable Newsletter](https://www.evolvable.site/newsletter) to slowly learn more, or keep reading this contextualization of the [full documentation](https://rubydoc.info/github/mattruzicka/evolvable).
 
-### The Evolvable Abstraction
-Population objects are composed of instances that include the `Evolvable` module. Instances are composed of gene objects that include the `Evolvable::Gene` module. Evaluation and evolution objects are used by population objects to evolve your instances. An evaluation object has one goal object and the evolution object is composed of selection, crossover, and mutation objects by default.
 
-All classes exposed by Evolvable are prefixed with `Evolvable::` and can be configured, inherited, removed, and extended. You can also choose between various Evolvable implementations documented below or substitute your own. Don't worry if none of this made sense, this will become clearer as you continue reading.
+## Table of Contents
+* [Installation](#installation)
+* [Getting Started](#getting-started)
+* [Concepts](#concepts)
+* [Genes](#genes)
+* [Populations](#populations)
+* [Evaluation](#evaluation)
+* [Evolution](#evolution)
+* [Selection](#selection)
+* [Combination](#combination)
+* [Mutation](#mutation)
+* [Search Space](#search-space)
+
 
 ## Installation
 
-Add `gem 'evolvable'` to your application's Gemfile and run `bundle install` or install it yourself with `gem install evolvable`
+Add [gem "evolvable"](https://rubygems.org/gems/evolvable) to your Gemfile and run `bundle install` or install it yourself with: `gem install evolvable`
 
 ## Getting Started
 
-After installing and requiring the "evolvable" Ruby gem:
+The `Evolvable` module makes it possible to implement evolutionary behaviors for
+any class by defining a `.search_space` class method and `#value` instance method.
+Then to evolve instances, initialize a population with `.new_population` and invoke
+the `#evolve` method on the resulting population object.
 
-1. Include the `Evolvable` module in the class for the instances you want to evolve. (See [Configuration](#Configuration)).
-2. Implement `.gene_space`, define any gene classes referenced by it, and include the `Evolvable::Gene` module for each. (See [Genes](#Genes)).
-3. Implement `#value`. (See [Evaluation](#Evaluation)).
-4. Initialize a population and start evolving. (See [Populations](#Populations)).
+### Implementation Steps
 
-Visit the [Evolvable Strings Tutorial](https://github.com/mattruzicka/evolvable/wiki/Evolvable-Strings-Tutorial) to see these steps in action. It walks through a simplified implementation of the [evolve string](https://github.com/mattruzicka/evolve_string) command-line program. Here's the [example source code](https://github.com/mattruzicka/evolvable/blob/master/examples/evolvable_string.rb) for the tutorial.
+1. [Include the `Evolvable` module in the class you want to evolve.](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable)
+2. [Define `.search_space` and any gene classes that you reference.](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/SearchSpace)
+3. [Define `#value`.](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Evaluation)
+4. [Initialize a population with `.new_population` and use `#evolve`.](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Population)
 
-If you’d like to quickly play around with an evolvable string Population object, you can do so by cloning this repo and running the command `bin/console` in this project's directory.
 
-## Usage
-- [Configuration](#Configuration)
-- [Genes](#Genes)
-- [Evaluation](#Evaluation)
-- [Populations](#Populations)
-- [Evolution](#Evolution)
-- [Selection](#Selection)
-- [Crossover](#Crossover)
-- [Mutation](#Mutation)
+To demonstrate these steps, we'll look at the [Hello World](#) example program.
 
-## Configuration
+### Hello World
 
-You'll need to define a class for the instances you want to evolve and include the `Evolvable` module. Let's say you want to evolve a melody. You might do something like this:
+Let's build the evolvable hello world program using the above steps. It'll evolve a population of arbitrary strings to be more like a given target string. After installing this gem, run `evolvable hello` at the command line to see it in action.
 
-```ruby
-class Melody
-  include Evolvable
-
-  def self.gene_space
-    # Expected
-  end
+Below is example output from evolving a population of randomly initialized string objects to match "Hello World!", then "Hello Evolvable World".
 
-  def value
-    # Required
-  end
-end
+```
+❯ Enter a string to evolve: Hello World!
+
+pp`W^jXG'_N`%              Generation 0
+H-OQXZ\a~{H*               Generation 1 ...
+HRv9X WorlNi               Generation 50 ...
+HRl6W World#               Generation 100 ...
+Hello World!               Generation 165
+
+❯ Enter a string to evolve: Hello Evolvable World
+
+Helgo World!b+=1}3         Generation 165
+Helgo Worlv!}:c(SoV        Generation 166
+Helgo WorlvsC`X(Joqs       Generation 167
+Helgo WorlvsC`X(So1RE      Generation 168 ...
+Hello Evolv#"l{ Wor*5      Generation 300 ...
+Hello Evolvable World      Generation 388
 ```
 
-The `Evolvable` module exposes the following class and instance methods for you.
-
-#### .gene_space
-
-You're expected to override this and return a gene space configuration hash or GeneSpace object.
+### Step 1
 
-Here's a sample definition for the melody class defined above. It configures each instance to have 16 note genes and 1 instrument gene.
+Let's begin by defining a `HelloWorld` class and have it **include the `Evolvable` module**.
 
 ```ruby
-def self.gene_space
-  { instrument: { type: 'InstrumentGene', count: 1 },
-    notes: { type: 'NoteGene', count: 16 } }
+class HelloWorld
+  include Evolvable
 end
 ```
 
-See the section on [Genes](#genes) for more details.
-
-#### .new_population(keyword_args = {})
+### Step 2
 
-Initializes a new population. Example: `population = Melody.new_population(size: 100)`
+Now we can **define the `.search_space`** class method with the types of [genes](#genes) that we want our our evolvable "hello world" instances to be able to have. We'll use `CharGene` instances to represent single characters within strings. So an instance with the string value of "Hello" would be composed of five `CharGene` instances.
 
-Accepts the same arguments as `Population.new`
+```ruby
+class HelloWorld
+  include Evolvable
 
-#### .new_instance(population: nil, genes: [], population_index: nil)
+  def self.search_space
+    ["CharGene", 1..40]
+  end
+end
+```
 
-Initializes a new instance. Accepts a population object, an array of gene objects, and the instance's population index.
+The [Search Space](#search-space) can be defined in a variety of ways. The above is shorthand that's useful for when there's only one type of gene. This method can also return an array of arrays or hash.
 
-This method is useful for re-initializing instances and populations that have been saved.
+The `1..40` specifies the range of possible genes for a particular HelloWorld instance. Evolvable translates this range or integer value into a `Evolvable::CountGene` object.
 
-_It is not recommended that you override this method_ as it is used by Evolvable internals. If you need to customize how your instances are initialized you can override either of the following two "initialize_instance" methods.
+By specifying a range, an `Evolvable::CountGene` instance can change the number of genes that are present in an evovlable instance. Count genes undergo evolutionary operations like any other gene. Their effects can be seen in the letter changes from Generation 165 to 168 in the above example output.
 
-#### .initialize_instance
+To finish step 2, we'll **define the gene class** that we referenced in the above `.search_space` method. Gene classes should include the `Evolvable::Gene` module.
 
-The default implementation simply delegates to `.new` and is useful for instances with custom initialize methods.
+```ruby
+class CharGene
+  include Evolvable::Gene
 
-#### #initialize_instance
+  def self.chars
+    @chars ||= 32.upto(126).map(&:chr)
+  end
 
-Runs after Evolvable finishes building your instance. It's useful for stuff like implementing custom gene initialization logic. For example, the Evolvable Strings web demo (coming soon) uses it to read from a "length gene" and add or remove "char genes" accordingly.
+  def to_s
+    @to_s ||= self.class.chars.sample
+  end
+end
+```
 
-#### #population, #population=
+It's important that, once accessed, the data for a particular gene never change. When the `#to_s` method first runs, Ruby's `||=` operator memoizes the result of randomly picking a char, enabling this method to sample a char only once per gene.
 
-The population object being used to evolve this instance.
+After defining the search space, we can now initialize `HelloWorld` instances with random genes, but to actually evolve them, we need to **define the `#value` instance method**. It provides the basis for comparing different evolvable instances.
 
-#### #genes, #genes=
+### Step 3
 
-An array of all an instance's genes. You can find specific types of genes with the following two methods.
+In the next step, we'll set the goal value to 0, so that evolution favors evolvable HelloWorld instances with `#value` methods that return numbers closer to 0. That means we want instances that more closely match their targets to return scores nearer to 0. As an example, if our target is "hello world", an instance that returns "jello world" would have a value of 1 and "hello world" would have a value of 0.
 
-#### #find_genes(key)
+For a working implementation, see the `#value` method in [examples/hello_world.rb](https://github.com/mattruzicka/evolvable/blob/main/examples/hello_world.rb)
 
-Returns an array of genes that have the given key. Gene keys are defined in the [.gene_space](####.gene_space) method. In the Melody example above, the key for the note genes would be `:notes`. The following would return an array of them: `note_genes = melody.find_genes(:notes)`
+### Step 4
 
-#### #find_gene(key)
+Now it's time to **initialize a population with `.new_population`**. By default, evolvable populations seek to maximize numeric values. In this program, we always know the best possible value, so setting the goal to a concrete number makes sense. This is done by passing the evaluation params with equalize set to 0.
 
-Returns the first gene with the given key. In the Melody example above, the instrument gene has the key `:instrument` so we might write something like: `instrument_gene = melody.find_gene(instrument)`
+We'll also specify the number of instances in a population using the population's `size` parameter and change the mutation porbability from 0.03 (3%) to 0.6 (60%).
 
-#### #population_index, #population_index=
+Experimentation has suggested that a large mutation probability tends to decrease the time it takes to evolve matches with short strings and has the opposite effect for long strings. This is demonstrated in the example output above by how many generations it took to go from "Hello World!" to "Hello Evolvable World". As an optimization, we could dynamically change the mutation probability using a population hook detailed below, but doing so will be left as an exercise for the reader. [Pull requests are welcome.](#contributing)
 
-Returns an instance's population index - an integer representing the order in which it was initialized in a population. It's the most basic way to distinguish instances in a population.
+```ruby
+population = HelloWorld.new_population(size: 100,
+                                       evaluation: { equalize: 0 },
+                                       mutation: { probability: 0.6 }
+```
 
-#### #value
 
-It is required that you implement this method. It is used when evaluating instances before undergoing evolution.
+At this point, everything should work when we run `population.evolve`, but it'll look like nothing is happening. The next section will allow us to gain instight by hooking into the evolutionary process.
 
-Technically, this method can return any object that implements Ruby's [Comparable](https://ruby-doc.org/core-2.7.1/Comparable.html). See the section on [Evaluation](#Evaluation) for details.
+### Evolvable Population Hooks
 
-#### Evolvable Hooks
+The following class methods can be implemented on your Evolvable class, e.g. HelloWorld, to hook into the Population#evolve lifecycle. This is useful for logging evolutionary progress, optimizing parameters, and creating interactions with and between evolvable instances.
 
-The following class methods can be overridden in order to hook into a population's evolutions. The hooks run for each evolution in the following order:
+1. `.before_evaluation(population)` - Runs before evaluation.
 
-**.before_evaluation(population)**
+2. `.before_evolution(population)`- Runs after evaluation and before evolution.
 
-**.before_evolution(population)**
+3. `.after_evolution(population)` - Runs after evolution.
 
-**.after_evolution(population)**
 
-To use our Melody example from above, you could override the `.before_evolution` method to play the best melody from each generation with something like this:
+Let's define `.before_evolution` to print the best value for each generation. We'll also define `HelloWorld#to_s`, which implicitly delegates to `CharGene#to_s` during the string interpolation that happens.
 
 ```ruby
-class Melody
+class HelloWorld
+  include Evolvable
+
   def self.before_evolution(population)
-    best_melody = population.best_instance
-    best_melody.play
+    best_evolvable = population.best_evolvable
+    evolutions_count = population.evolutions_count
+    puts "#{best_evolvable} - Generation #{evolutions_count}"
   end
 
-  def play
-    note_genes = melody.find_genes(:notes)
-    note_values = note_genes.map(&:value)
-    find_gene(:instrument).play(note_values)
+  # ...
+
+  def to_s
+    @to_s ||= genes.join
   end
+
+  # ...
 end
 ```
 
-## Genes
+Finally we can **evolve the population with the `Evolvable::Population#evolve` instance method**.
 
-Instances rely on gene objects to compose behaviors. In other words, a gene can be thought of as an object that in some way affects the behavior of an instance. They are used to encapsulate a "sample space" and return a sample outcome when accessed.
+```ruby
+population.evolve
+```
 
-#### The Evolvable::Gene module
+**You now know the fundamental steps to building evolvable programs of endless complexity in any domain!** 🐸 The exact implementation for the command line demo can be found in [exe/hello](https://github.com/mattruzicka/evolvable/blob/main/exe/hello) and [examples/hello_world.rb](https://github.com/mattruzicka/evolvable/blob/main/examples/hello_world.rb).
 
-Gene objects must include the `Evolvable::Gene` module which enables them to undergo evolutionary operations such as crossover and mutation.
+## Concepts
 
-To continue with the melody example, we might encode a NoteGene like so:
+[Populations](#populations) are composed of evolvables which are composed of genes. Evolvables orchestrate behaviors by delegating to gene objects. Collections of genes are organized into genomes and constitute the [search space](#search-space). [Evaluation](#evaluation) and [evolution](#evolution) objects are used to evolve populations. By default, evolution is composed of [selection](#selection), [combination](#combination), and [mutation](#mutation).
 
-```ruby
-class NoteGene
-  include Evolvable::Gene
+The following concept map depicts how genes flow through populations.
 
-  NOTES = ['C', 'C♯', 'D', 'D♯', 'E', 'F', 'F♯', 'G', 'G♯', 'A', 'A♯', 'B']
+![Concept Map](https://github.com/mattruzicka/evolvable/raw/main/examples/images/diagram.png)
 
-  def value
-    @value ||= NOTES.sample
-  end
-end
-```
-Here, the "sample space" for the NoteGene class has twelve notes, but each individual object will have only one note. The note is randomly chosen when the "value" method is invoked for the first time.
+Evolvable is designed with extensibility in mind. Evolvable objects such as [evaluation](#evaluation), [evolution](#evolution), [selection](#selection), [combination](#combination), and [mutation](#mutation) can be extended and swapped, potentially in ways that alter the above graph.
+
+## Genes
+For evolution to be effective, an evolvable's genes must be able to influence
+its behavior. Evolvables are composed of genes that can be used to run simple
+functions or orchestrate complex interactions. The level of abstraction is up
+to you.
 
-_It is important that the data for a particular gene never change._ Ruby's or-equals operator `||=` is super useful for memoizing gene attributes like this. It is used above to randomly pick a note only once and return the same note for the lifetime of the object.
+Defining gene classes requires encapsulating some "sample space" and returning
+a sample outcome when a gene attribute is accessed. For evolution to proceed
+in a non-random way, the same sample outcome should be returned every time
+a particular gene is accessed with a particular set of parameters.
+Memoization is a useful technique for doing just this. The
+[memo_wise](https://github.com/panorama-ed/memo_wise) gem may be useful for
+more complex memoizations.
 
-A melody instance with multiple note genes might use the `NoteGene#value` method to compose the notes of its melody like so: `melody.find_genes(:note).map(&:value)`. Let's keep going with this example and implement the `InstrumentGene` too:
 
 ```ruby
-class InstrumentGene
-  include Evolvable::Gene
+# This gene generates a random hexidecimal color code for use by evolvables.
 
-  def instrument_class
-    @instrument_class ||= [Guitar, Synth, Trumpet].sample
-  end
+require 'securerandom'
 
-  def volume
-    @volume ||= rand(1..100)
-  end
+class ColorGene
+  include Evolvable::Gene
 
-  def play(notes)
-    instrument_class.play(notes: notes, volume: volume)
+  def hex_code
+    @hex_code ||= SecureRandom.hex(3)
   end
 end
 ```
 
-You can model your sample space however you like. Ruby's [Array](https://ruby-doc.org/core-2.7.1/Array.html), [Hash](https://ruby-doc.org/core-2.7.1/Hash.html), [Range](https://ruby-doc.org/core-2.7.1/Range.html), and [Random](https://ruby-doc.org/core-2.7.1/Random.html) classes may be useful. This InstrumentGene implementation has 300 possible outcomes (3 instruments * 100 volumes) and uses Ruby's Array, Range, and Random classes.
 
-Now that its genes are implemented, a melody instance can use them:
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Gene)
 
-```ruby
-class Melody
-  include Evolvable
+## Populations
+Population objects are responsible for generating and evolving instances.
+They orchestrate all the other Evolvable objects to do so.
 
-  # ...
+Populations can be initialized and re-initialized with a number of useful
+parameters.
 
-  def play
-    note_genes = melody.find_genes(:notes)
-    note_values = note_genes.map(&:value)
-    find_gene(:instrument).play(note_values)
-  end
-end
-```
 
-In this way, instances can express behaviors via genes and even orchestrate interactions between them. Genes can also interact with each other during an instance's initialization process via the [#initialize_instance](#initialize_instance-1) method
+```ruby
+# TODO: initialize a population with all supported parameters
+```
 
-#### The Evolvable::GeneSpace object
 
-TODO
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Population)
 
 ## Evaluation
+For selection to be effective in the context of evolution, there needs to be
+a way to compare evolvables. In the genetic algorithm, this is often
+referred to as the "fitness function".
 
-TODO
+The `Evolvable::Evaluation` object expects evolvable instances to define a `#value` method that
+returns some numeric value. Values are used to evaluate instances relative to each
+other and with regards to some goal. Out of the box, the goal can be set
+to maximize, minimize, or equalize numeric values.
 
-#### The Evolvable::Evaluation object
 
-TODO
+```ruby
+# TODO: Show how to add/change population's evaluation object
 
-#### The Evolvable::Goal::Maximize object
+# The goal value can also be assigned via as argument to `Evolvable::Population#evolve`
+population.evolve(goal_value: 1000)
+```
 
-TODO
 
-#### The Evolvable::Goal::Minimize object
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Evaluation)
 
-TODO
+## Evolution
+After a population's instances are evaluated, they undergo evolution.
+The default evolution object is composed of selection,
+crossover, and mutation objects and applies them as operations to
+a population's evolvables in that order.
 
-#### The Evolvable::Goal::Equalize object
 
-TODO
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Evolution)
 
-## Populations
+## Selection
+The selection object assumes that a population's evolvables have already
+been sorted by the evaluation object. It selects "parent" evolvables to
+undergo combination and thereby produce the next generation of evolvables.
 
-TODO
+Only two evolvables are selected as parents for each generation by default.
+The selection size is configurable.
 
-#### The Evolvable::Population object
 
-TODO
+```ruby
+# TODO: Show how to add/change population's selection object
+```
 
-## Evolution
 
-TODO
 
-#### The Evolvable::Evolution object
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Selection)
 
-TODO
+## Combination
+Combination generates new evolvable instances by combining the genes of selected instances.
+You can think of it as a mixing of parent genes from one generation to
+produce the next generation.
 
-## Selection
+You may choose from a selection of combination objects or implement your own.
+The default combination object is `Evolvable::GeneCombination`.
 
-TODO
 
-#### The Evolvable::Selection object
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Combination)
 
-TODO
+## Mutation
+Mutation serves the role of increasing genetic variation. When an evolvable
+undergoes a mutation, one or more of its genes are replaced by newly
+initialized ones. In effect, a gene mutation invokes a new random outcome
+from the genetic search space.
 
-## Crossover
+Mutation frequency is configurable using the `probability` and `rate`
+parameters.
 
-TODO
 
-#### The Evolvable::GeneCrossover object
+```ruby
+# Show how to initialize/assign population with a specific mutation object
+```
 
-TODO
 
-#### The Evolvable::UniformCrossover object
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/Mutation)
 
-TODO
+## Search Space
+The search space encapsulates the range of possible genes
+for a particular evolvable. You can think of it as the boundaries of
+genetic variation. It is configured via the
+[.search_space](#evolvableclasssearch_space) method that you define
+on your evolvable class. It's used by populations to initialize
+new evolvables.
 
-#### The Evolvable::PointCrossover object
+Evolvable provides flexibility in how you define your search space.
+The below example implementations for `.search_space` produce the
+exact same search space for the
+[Hello World](https://github.com/mattruzicka/evolvable#hello-world)
+demo program. The different styles arguably vary in suitability for
+different contexts, perhaps depending on how programs are loaded and
+the number of different gene types.
 
-TODO
 
-## Mutation
+```ruby
+# All 9 of these example definitions are equivalent
+
+# Hash syntax
+{ chars: { type: 'CharGene', max_count: 100 } }
+{ chars: { type: 'CharGene', min_count: 1, max_count: 100 } }
+{ chars: { type: 'CharGene', count: 1..100 } }
+
+# Array of arrays syntax
+[[:chars, 'CharGene', 1..100]]
+[['chars', 'CharGene',  1..100]]
+[['CharGene', 1..100]]
+
+# A single array works when there's only one type of gene
+['CharGene', 1..100]
+[:chars, 'CharGene', 1..100]
+['chars', 'CharGene', 1..100]
+```
+
+
+[Documentation](https://rubydoc.info/github/mattruzicka/evolvable/Evolvable/SearchSpace)
 
-TODO
+## Contributing
 
-#### The Evolvable::Mutation object
+Bug reports and pull requests are welcome on GitHub at https://github.com/mattruzicka/evolvable.
 
-TODO
+If you're interested in contributing, but don't know where to get started, message me on twitter at [@mattruzicka](https://twitter.com/mattruzicka).