3 ms·
Isn't data and processing power the most important thing with neural networks? Even if I knew how they worked I would have no idea what to do with them that has
by Question1101 10y ago
Isn't data and processing power the most important thing with neural networks? Even if I knew how they worked I would have no idea what to do with them that hasn't been done already as a hobbyst without access to huge amounts of data like companies do.
- eegilbert 10y agoYes. This is often, perhaps willfully given the incentives, overlooked by computer scientists (and people trained in that discipline). For a complete, cogent argument see: http://static.googleusercontent.com/media/research.google.com/en//pubs/archive/35179.pdf http://static.googleusercontent.com/media/research.google.co...
- naveen99 10y agoData is cheap and all around us. Processing power yes, but you can get started with a gaming pc.
- p1esk 10y agoNo. As a researcher, you can make it your goal to find/invent a smallest possible architecture for a given task (in terms of number parameters, or number of operations). Alternatively, you can try to invent an architecture to learn faster from data (or require less data to achieve state of the art results).
- tree_of_item 10y ago> As a researcher But the post you replied to specifically said "as a hobbyist", so it doesn't really sound like there's much hope.
- Joof 10y agoHobbyist researcher? Seems like a more accessible plan anyway; fuck bothering with huge datasets and long training times and just focus on optimizing small architecture.
- discardorama 10y agoIf you want to play around, take a gander at Kaggle.
- nl 10y agoYes, you need data. But the amount depends on your task, and there are pretty significant sources of large amounts of data available online. For example I was at a presentation where a person built a pretty interesting neural model based on 190k clinical records released via Kaggle. In most fields it is surprising how much data is easily accessible.