4 ms·
Then most likely they would not be successful in their businesses; training older models or newer ones with a batch of 1 is not a recipe for success these days
by bitL 8y ago
Then most likely they would not be successful in their businesses; training older models or newer ones with a batch of 1 is not a recipe for success these days when you more like need 100s of GPUs running slight modifications to models in parallel to find the one that works.
- sgt101 8y agoWe used to call this model fishing and regard models that came from places that did it with fear and suspicion (due to sudden failure and poor performance in production) What has changed that people think this is a wise approach?
- bitL 8y agoAutoML and automated architecture/hyperparameter search used by FAANG, producing things like AmoebaNet-D, that beat all previous state-of-art results.
- matt4077 8y agoThey don't suddenly fail any more, and performance in production is fine... This is a rather empirical science right now. If forced to speculate, I'd say far larger datasets are part of the answer. If you can afford to hold back 30%+ of your data for (multi-tranche) verification the difference between testing and production becomes a problem for the philosophers.
- sgt101 8y agoHmm this isn't my experience. My experience is that highly complex models tend to reflect accurately the ground truth in the current dataset, but not the domain theory that generated it, so when things change (like, prevailing weather conditions, political alignments, interest rates, data usage) which move the distribution (but not the domain theory) they do fail. The question is : are you trading poorer compression for more fidelity or are you learning the system that generates the data?
- sabalaba 8y agoMost people I know who publish in NIPS/ICLR train their models using between 1-4 GPUs under their desk or 8 GPUs in a server. I would argue that these folks make competitive models and "are successful in their business" with only 1-8 GPUs. While having access to hundreds of GPUs helps you explore model space more quickly, it's not a hard pre-requisite to success in Deep Learning.
- eanzenberg 8y agoYup, state of the art doesn't require GPU farms, just a couple in a consumer desktop will do.
- bitL 8y agoLookup AmoebaNet-D if you think you don't need a GPU farm these days. Those were times when 1-2 GPUs were enough...
- jefft255 8y agoYou know, there is more to deep learning research than meta-learning/architecture exploration. Sure you can explore the hyper-parameter space faster with 500 GPUs and get yet again a 0.05% better test score on ImageNet (or more I don't actually know), but there are other ways to do something meaningful in DL without using such compute power.
- bitL 8y agoThat's a fair point and I agree. It's just sometimes difficult to beat automated exploration; as a standard company you probably don't have access to top-end researchers/practitioners, just average ones, and those might get a significant boost by trading smartness for brute force and run many models in parallel in evolutionary fashion. When you e.g. look at how Google's internal system constructs loss functions and how many combinations do they try, one has to have an unexpected idea to beat their results, and that idea can be usually quickly incorporated into their platform, raising the bar for individual researchers. At Facebook they basically press a few buttons and select a few checkboxes, then wait until best model is selected, leading to frustration among researchers.