Federated ML to cut the mustard for privacy concerned industries

Oana Olteanu
Mar 23 · 2 min read
https://www.dictionary.com/e/cut-the-mustard/

Federated learning was invented at Google in 2015 and represents a better way to do ML at the edge. Recently it moved from research to being applied and useful. What it means is, for each device, a model is trained with the data available to that device, and then each of the models sends back some parameters to the server. Basically, the data does not move, it remains on the device, only the parameters are used to train a model from all the nodes that reported back. Once the master model gets better, it pushes updates to the devices, making each individual device model better. In other words, federated learning is distributed ML across devices.

Federated learning is important because it frees companies from moving data, which means less engineering trouble and less privacy trouble. It is also important because it allows for more accurate ML since the model can be trained on a higher, more varied volume of data. One area where this has proven to lead to better results is in predictive maintenance. Healthcare could also benefit from it greatly and financial services. These industries are handling valuable PII data that must be protected at all costs. The custodians of this data have to choose between privacy OR getting more value out of the data. Federated learning would transform the OR statement into an and statement, privacy AND making use of the data, without exposing the data for mishandling or fraud.

Federated Learning is happening at the same time as WebAssembly and Rust. This makes for a killer combination. Rust is a better C++ with a static garbage collector for better memory management. With Rust, developers can write safer code for powerful applications. WebAssembly moves these powerful apps to the browser by giving them a near-native feel and speed. When we use these apps in the browser, we generate a lot of data that is personal and must be protected. Federated Learning preserves the privacy of the data and enables value creation without compromising us and our data. Federated Learning + WebAssembly + Rust = safer, faster, more valuable apps #futureisbright

If you are a federated learning practitioner, I’d love to grab a virtual coffee and share learnings.

Oana

Oana Olteanu

Written by

I like to invest in and partner with founders to help them scale their ventures. I have a degree in and a soft spot for ML @oanaolt

More From Medium

Also tagged Rust

Also tagged AI

Also tagged Webassembly

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade