Jadezzz / DL-Adder-Subtractor

Using Deep Learning to implement adder and subtractor

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

DL-Adder-Subtractor

Notes for TAs

For Coding part, please refer to Demo.ipynb in this repo or this nbviewer link

For Analyze part, please refer to Report.md in this repo

Introduction

使用深度學習的技術(主要爲 Recurrent Neural Network),讓電腦學習:

  1. 加法 (3~5位數)
  2. 減法(3~5位數)
  3. 加法與減法合並(3~5 位數)

Flow

Model

Encoder 的部分,可以看成先將待求的算式理解並且歸納成爲一個向量(128維度)

Decoder 的部分,則是通過 Encoder 的向量輸出,通過 LSTM 之後,輸出128 維度向量,經過 12 個 neurons 後再通過 soft-max 預測該位置的數字

(注:以 1 層 LSTM Decoder,Output 4 位數爲例)

About

Using Deep Learning to implement adder and subtractor


Languages

Language:Jupyter Notebook 88.0%Language:Python 12.0%