4 ms·
I think you could be spot on: there are new applications emerging in deep learning, like self-driving vehicles, where you have a powerful need for mountains of
by jd20 10y ago
I think you could be spot on: there are new applications emerging in deep learning, like self-driving vehicles, where you have a powerful need for mountains of data to train complex models, yet a logistics problem in how to aggregate that data in a single place (I'm making this up, but imagine a car collecting 7 streams of 4K video at 60 fps). I really see a growing need for these types of distributed training models.
- Fricken 10y agoI'm a bit confused. Doesn't the data need to be cleaned up, annotated, or labelled before it can be used to train? How would this work if the data doesn't leave the local device?
- vitohuang 10y agoIt did mentioned a local version of mini TensorFlow, I would guess the data is cleaned up on local devices and only updates are send back to the cloud.
- jd20 10y agoSounded like the label for training is the user's action (whether they chose the suggestion provided or not). Then they collect the model updates in bulk from millions of users, that's far more useful for training purposes than a small very well curated data set.
- amelius 10y agoThere is a branch called unsupervised learning, where no labeling is needed.
- ska 10y agoThere is, but supervised and unsupervised learning mostly don't solve the same problems - so this doesn't necessarily help
- bryanrasmussen 10y agoas I understand what I read the data does leave, but just the users updates - sort of like only your changes to a git a branch are sent when you push.
- XJOKOLAT 10y agoNot knowing enough about this, my first impression is: How does Federated Learning cope with ... FAKE NEWS! (for example).
- brk 10y agoIt is mostly applicable for taking data sourced at an endpoint (e.g. Mobile phone) and running what is essentially a refinement to the learning. A key component is that to analyze the data in the cloud for the same refinement would mean sending the data to the cloud, which the user may not want, and may also be bandwidth intensive. For fake news, the data is already in the cloud, being pushed down to the users device. A user could mark something as 'fake' (via explicit action, or possibly inaction), and that 'mark' is uploaded and the data is analyzed by central compute for refinement. To be clear, that is NOT what this paper is about, I am saying fake news would be a bad use, because the data is not being sourced by the user, only viewed and possibly marked. A better example than the gboard gesture learning refinements might be in the form of an app that acts as a dashcam (leaving aside the kludgeyness of that). The user could mark things like proper recognition of street signs, traffic lights, or brands/models of vehicles. The phone would then analyze those classification, compute a diff to the algorithm, send just that diff to the cloud (summarizing a lot). Multiply that by 1,000,000 users, and now you have a refined data set that did not require sending 100,000,000,000 images to the cloud for analysis.