Machine Learning for Mobile apps
What is the device, almost everyone in this world are engaging most of the time?
Yes. Its “Mobile phone”. Like other electronic devices like laptops/tablets, the engagement rate on mobile is increasing day by day.
Why ML/AI create potential in Mobile Apps?
Camera & Photos
The best example is “Camera” in Mobile keeps on getting added with new features. The year 2019’s launched with “AI-powered camera mobiles” depicts the same. “Camera” remains top of mind for a buyer and became the primary selling point for a newly launching mobile phone. Many apps also are released based on Camera and Photos like “Lensa” — Prisma’s new AI-powered app. AI photography adjusts the colors and photoshop elements based on the subject by itself.
Space and Memory
Also, we have passed the days when most were worrying about space and memory to install new apps and store things in mobiles. I really like the first line of this post: “Today we have more computing power in a single smartphone than the entire NASA space program used to put a man on the moon.”
Users expectations are /sky-high
Just imagine the difference. An email app that supports only basic email operations. And another email app e.g. Gmail that does the same work differently like automatically segregating incoming mails as Primary, Social and Promotions. Just think, which one will have a hit? Now, let us see different ML algorithms/patterns that operate behind this:
- Spam or Not spam — Binary classification algorithm,
- Personal, Promotion, Social — K-means clustering algorithm,
- Mail content suggestions — Text Analytics
- In addition, we also see auto-suggestions while writing mail, alerts about bills overdue, alerts about mails promised to reply but we didn’t do and the list keep-on increasing.
Mobile-AI in future — “AI Edge”
AI will be on the mobile device itself i.e. if ML model library that computes the data is in your app itself, then it makes more sense coming to privacy, security, and latency. Also, it will work offline without internet. Still, it has some disadvantages like intensive memory consumption and battery of mobile at times to compute the result.
Pre-built AI — Microsoft Cognitive Services
There are many pre-built AI services for Vision, Text, etc. like Microsoft’s Cognitive. You can directly use such services via REST API in any application model. The respective data will be sent to the cloud, computed and the result brought back to mobile.
ML Frameworks
Let us see some frameworks that are used to incorporate artificial intelligence — machine learning features into mobile apps.
Core ML
“Core ML” is a famous ML Framework for Mac devices. Core ML stays at its best than other frameworks for IOS; because of its super simple to run pre-trained models on Apple devices. It is built on top of previously built ML frameworks Accelerate and BNNS. Core ML is an abstraction on top of these two.
ML.Net
“ML.NET” is an open-source and cross-platform machine learning framework. Creating your own model will be a super-duper easy with Auto ML if you are already a .Net developer. Telling this from my own experience. You can deploy the model in a web app and consume it wherever needed.
Over to you
Programming languages would vary. But the principles remain the same. If one is strong enough in the basics of ML and has the thrive to solve real problems of users, he/she can create a stunning AI-powered mobile app.
Thanks for your time reading my blog until the end!
More references,
- Check out my blog about artificial intelligence in the medium as well.
- Here, you can find more blogs about app analytics.
See you again soon :)
