Archive

# DeepThinkJS JS Machine Learning Library!

DeepThinkJS is an open-source artifical ingelligence library designed for programmers to quickly implement AI algorithms that are already build in.

The purpose:

As you might already know, nowadays, JS has became a very popular language that can interact with html5 very easily, thus having the ability to illustrate

However, when I first try to find AI library for JS, there is only a few for NodeJS, so I decided to build this library to help programmers quickly implement AI algorithms that are already pre-programmed in the library, so they do not have to spend hours and hours trying to create the model

OKAY~ Enough with the overview, lets get to the specific details about this library

Supported Algorithms: MLP(Multilayer Perceptron) kNN(k-nearest-neighbours) CNN(Convlutional Neural Network)

Supported Activation Functions: Sigmoid Tanh ReLU

Library Size: 11.5KB (pretty light I would say)

Difficulty to use: Pretty Easy (You just have to set the parameter and feed data in it, then you get the output)

Status: There are still bugs and missing common algorithms, but I am still working it, so I think I am able to expand the library over time.

Supported Language: JS

Develops: CodeTiger(only me o(╥﹏╥)o)

Note1: The Convolutional Neural Network is only partially functional right now, the backprogation still has bugs.

Note2: For Multilayey Perceptron, ReLU is not working properly.

As you can see, there are still a lot of improvements that can be made (⊙︿⊙). However, I am constantly working it, so the library is getting better and better (*^▽^*)

Now for the DEMO ✿✿ヽ(°▽°)ノ✿.

the demo below is basically the artificial intelligence trying to learn to match the pattern of the function, which is x * x / 5 - x - 2 * Math.abs(Math.sin(x * 2), which is basically a function with a lot of curves

Tip: If the red line is already matched with the white line, then try refreshing.

Finally, the source code for this project is in github, the link is https://github.com/tiger927/DeepThinkJS

The purpose:

As you might already know, nowadays, JS has became a very popular language that can interact with html5 very easily, thus having the ability to illustrate

However, when I first try to find AI library for JS, there is only a few for NodeJS, so I decided to build this library to help programmers quickly implement AI algorithms that are already pre-programmed in the library, so they do not have to spend hours and hours trying to create the model

OKAY~ Enough with the overview, lets get to the specific details about this library

Supported Algorithms: MLP(Multilayer Perceptron) kNN(k-nearest-neighbours) CNN(Convlutional Neural Network)

Supported Activation Functions: Sigmoid Tanh ReLU

Library Size: 11.5KB (pretty light I would say)

Difficulty to use: Pretty Easy (You just have to set the parameter and feed data in it, then you get the output)

Status: There are still bugs and missing common algorithms, but I am still working it, so I think I am able to expand the library over time.

Supported Language: JS

Develops: CodeTiger(only me o(╥﹏╥)o)

Note1: The Convolutional Neural Network is only partially functional right now, the backprogation still has bugs.

Note2: For Multilayey Perceptron, ReLU is not working properly.

As you can see, there are still a lot of improvements that can be made (⊙︿⊙). However, I am constantly working it, so the library is getting better and better (*^▽^*)

Now for the DEMO ✿✿ヽ(°▽°)ノ✿.

the demo below is basically the artificial intelligence trying to learn to match the pattern of the function, which is x * x / 5 - x - 2 * Math.abs(Math.sin(x * 2), which is basically a function with a lot of curves

Tip: If the red line is already matched with the white line, then try refreshing.

Finally, the source code for this project is in github, the link is https://github.com/tiger927/DeepThinkJS

CodeTiger

2018.11.9