This shows you the differences between two versions of the page.
Both sides previous revision Previous revision Next revision | Previous revision | ||
research:largescale [2012/12/24 11:43] leonb [Stochastic Gradient for Large-Scale Learning] |
research:largescale [2013/02/25 09:57] (current) leonb [Papers] |
||
---|---|---|---|
Line 49: | Line 49: | ||
===== Approximate Optimization ===== | ===== Approximate Optimization ===== | ||
+ | {{ wall2.png}} | ||
Large-scale machine learning was first approached as an engineering problem. For instance, to leverage a | Large-scale machine learning was first approached as an engineering problem. For instance, to leverage a | ||
larger training set, we can use a parallel computer to run a known machine learning algorithm | larger training set, we can use a parallel computer to run a known machine learning algorithm | ||
Line 54: | Line 55: | ||
objective function. Such approaches rely on the appealing | objective function. Such approaches rely on the appealing | ||
assumption that one can decouple the statistical aspects from the computational aspects of the machine | assumption that one can decouple the statistical aspects from the computational aspects of the machine | ||
- | learning problem. My [[: | + | learning problem. |
- | that this assumption is incorrect, and that giving it up leads to considerably | + | |
- | more effective learning algorithms. | + | |
- | The [[: | + | This work shows that this assumption is incorrect, and that giving it up leads to considerably |
- | develops a theoretical framework | + | more effective learning algorithms. A new theoretical framework |
- | that takes into account the effect of approximate | + | takes into account the effect of approximate |
optimization on learning algorithms. | optimization on learning algorithms. | ||
+ | |||
The analysis shows distinct tradeoffs for the | The analysis shows distinct tradeoffs for the | ||
case of small-scale and large-scale learning problems. | case of small-scale and large-scale learning problems. | ||
Line 70: | Line 70: | ||
complexity of the underlying optimization | complexity of the underlying optimization | ||
algorithms in non-trivial ways. | algorithms in non-trivial ways. | ||
- | For instance, | + | For instance, [[: |
- | [[: | + | appear to be mediocre optimization algorithms and yet are shown to |
- | is shown to perform | + | [[: |
+ | |||
+ | |||
+ | |||
+ | ===== Tutorials ===== | ||
+ | |||
+ | * NIPS 2007 tutorial " | ||
+ | |||
+ | ===== Related ===== | ||
+ | |||
+ | * [[: | ||
+ | ===== Papers ===== | ||
+ | <box 99% orange> | ||
+ | Léon Bottou and Olivier Bousquet: | ||
+ | //Advances in Neural Information Processing Systems//, 20, | ||
+ | MIT Press, Cambridge, MA, 2008. | ||
+ | [[: | ||
+ | </ | ||
+ | <box 99% orange> | ||
+ | Léon Bottou and Yann LeCun: | ||
+ | [[: | ||
+ | </ | ||
+ | <box 99% orange> | ||
+ | Léon Bottou: | ||
+ | [[: | ||
+ | </ | ||
- | ===== Active Learning ===== | + | <box 99% blue> |
+ | Léon Bottou: | ||
- | One simple way to handle large-scale learning problems is to chose examples wisely. | + | [[: |
- | This idea was explored in our work on [[lasvm|Active and Online Support Vector Machines]]. | + | </ |
- | But there is still much work to do about active learning as a way to handle very large data repositories. | + | |