https://github.com/nedomas/future

Welcome to the future, really.
https://github.com/nedomas/future

Last synced: 19 days ago
JSON representation

Welcome to the future, really.

• Host: GitHub
• URL: https://github.com/nedomas/future
• Owner: Nedomas
• Created: 2018-01-12T17:54:48.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2018-01-13T09:30:51.000Z (almost 7 years ago)
• Last Synced: 2024-10-22T07:41:15.490Z (2 months ago)
• Language: Ruby
• Homepage:
• Size: 16.6 KB
• Stars: 0
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Future

*Welcome to the future, really.*

This is a proof-of-concept of what machine-learning-first programming language might look like.

I've implemented a future version of a `map` method in Ruby. It is usually done procedurially, but in the `Future` language it is done through a naive Bayes classifier. The classifier can be easily replaced by neural networks or some advanced tools like auto-ml.

Here's an example of how it works:

```ruby

descriptions = ['I code with laravel', 'I like it']

# this is a regular Ruby code you would write to map out

# whether the description matches text 'laravel'.

# What's different is that underneath it trained a naive Baynes classifier

laravel_description_network = descriptions.map do |description|

  !!description.match('laravel')

end

# => [true, false]

# Looks normal. But it actually got the result through the classifier and not procedurially.

# You can do the same for a title match:

titles = ['UI designer', 'php programmer']

php_title_network = titles.map do |title|

  !!title.match('php')

end

# => [false, true]

# Nothing special.

# Whats special is that you can now combine your already trained networks:

people = [

  { name: 'John', title: 'php dude', description: 'I must do laravel' },

  { name: 'Peter', title: 'non cool dude', description: 'I must do something about it' }

]

people_mapped_with_laravel = people.map([

  { network: laravel_description_network, transform: ->(person) { person[:description] } },

  { network: php_title_network, transform: ->(person) { person[:title] } }

]) do |person, combined_networks_result_works_with_laravel|

  { person[:name] => combined_networks_result_works_with_laravel }

end

# => [

  { 'John' => true },

  { 'Peter' => false }

]

# This `people.map` method combined the injected networks and trained the new combined

# network with it as inputs.

# `transform` key allows to transform the value of current `map` item to match to an already

# pretrained network.

# 

# So is the future?

```

## Why and how

Inspired by what the future of writing software might look like. 

Thanks to [Danielius](https://github.com/dvisockas) and [Mindaugas](https://github.com/kurmis) for exciting late-night talks about it.

## Installation

1. Clone this repository.

2. Install dependencies via:

```shell

$ bundle install

```

    

3. Run the test to see it in action:

```shell

$ ./bin/test_description_title

```

   

## Explanation

The usage is demonstrated in `lib/future.rb`