DEV Community

Cover image for List of Famous Machine Learning Patents
Hiromichi NOMATA
Hiromichi NOMATA

Posted on • Originally published at algonote.com

List of Famous Machine Learning Patents

Summary of famous machine learning patents

Tokyo patent office

Introduction

Machine learning community is relatively open. Papers can be dowloaded from arXiv for free. On the other hand, we sometimes found some methods are registered as patents.

I daily collect these patents shared on Twitter and Reddit. The patents below are the list of machine learning patents. Some patents are under application, not approved.

General

Generating output sequences from input sequences using neural networks by Google

http://www.freepatentsonline.com/10402719.html

known as seq2seq

Preventing overfitting

System and method for addressing overfitting in a neural network by Google

https://patents.google.com/patent/WO2014105866A1

known as Dropout

Batch normalization layers

https://patents.google.com/patent/US20160217368A1/en

As title shows, Batch Normalization patents. According to this slide, the patents are registered in many countries.

Image, Augmentation

Object detection and classification in images by Microsoft

http://patft.uspto.gov/netacgi/nph-Parser?Sect2=PTO1&Sect2=HITOFF&p=1&u=/netahtml/PTO/search-bool.html&r=1&f=G&l=50&d=PALL&RefSrch=yes&Query=PN/9858496

Faster R-CNN (object detection). According to qiita article, it is only applied to U.S.

DATA AUGMENTATION FOR IMAGE CLASSIFICATION TASKS by IBM

http://www.freepatentsonline.com/y2019/0087694.html

similar to Mixup

Sound

Spectrogram to waveform synthesis using convolutional networks by Baidu

https://patents.google.com/patent/US20190355347A1/en

waveform generation from spectrogram by (Multi-head) CNN

Machine learning to generate music from text by Google

https://pdfaiw.uspto.gov/.aiw?PageNum=0&docid=20180190249&IDKey=&HomeUrl=%2F

Music Generation from text

Others

Deep reinforcement learning for robotic manipulation by Google

https://patents.google.com/patent/WO2018053187A1/en

Multiple robots learn individually. By using acquired experiments, policy network is learned.

Active Machine Learning by Microsoft

https://patents.google.com/patent/US20160162802

When new labels are added, model start relearning in the limit of model size.

Impression

US patent office was open for comments about AI patents.

If it is paper, the explicit description like xx% precision improvement is popular. The patents description is ambiguous for enlarging the claim rights.

If you know the other patens, please let me know.

Top comments (0)