4 ms·

Popular Machine Learning algorithms

33 points by joydeep314 9y ago

sgt101 9y ago
You can state that a neural network is a hypothesis set, but is it useful when you've got a hypothesis that is 100's of values each with billions of possible values? I mean why talk about that in terms of a set?
- nonbel 9y ago
  From their link: >"Hypothesis set is the set of functions that have some common and unique properties that make them a viable candidate to be considered for the final hypothesis. We apply the learning algorithm to the hypothesis set and then choose a specific function from the hypothesis set. In terms of examples, the Hypothesis set could be linear regression and the learning algorithm could be gradient descent or the hypothesis set could be neural networks and the learning algorithm could be back propagation. Therefore, in order words, hypothesis set is a set of similar functions which have been shown to have good results under specific conditions and the learning algorithm is the algorithm that will do the actual searching." https://medium.com/technology-nineleaps/what-is-machine-learning-8f4f117a9f4e https://medium.com/technology-nineleaps/what-is-machine-lear... I agree with you, I don't find talking about "hypothesis sets" to be at all enlightening. To me it sounds like needless jargon and it is strange to see even this amount of effort expended to use it. Maybe we are missing something here though.
vosper 9y ago
This may be a very naïve question, but how far are we from toolkit that takes a dataset and a scoring function or goal and figures out which is the best technique, and optimizes it too? Sure, it's probably really resource intensive, but can it be done? Maybe with some kind of genetic algorithm selecting techniques and parameters? Soneone must be working on this.
- flatfilefan 9y ago
  An example of such tool https://github.com/paypal/autosklearn-zeroconf https://github.com/paypal/autosklearn-zeroconf There are many more including genetic algorithms. Some are discussed here: https://www.reddit.com/r/MachineLearning/comments/6efs8u/d_anyone_have_experience_with_automated_ml_tools/ https://www.reddit.com/r/MachineLearning/comments/6efs8u/d_a...
  - wakkaflokka 9y ago
    TPOT is also a good automl package that utilizes sklearn in Python. It's also actively under development. (I'm not affiliated, just an occasional user) https://github.com/rhiever/tpot https://github.com/rhiever/tpot
- cf 9y ago
  The keyword you are looking for is AutoML. http://automl.info/ http://automl.info/
- onuralp 9y ago
  You might also be interested to check out Automatic Statistician (developed by Z. Ghahramani, J. Tenenbaum and others) - https://www.automaticstatistician.com/index/ https://www.automaticstatistician.com/index/

33 points by joydeep314 9y ago

You can state that a neural network is a hypothesis set, but is it useful when you've got a hypothesis that is 100's of values each with billions of possible values? I mean why talk about that in terms of a set?

nonbel 9y ago
From their link: >"Hypothesis set is the set of functions that have some common and unique properties that make them a viable candidate to be considered for the final hypothesis. We apply the learning algorithm to the hypothesis set and then choose a specific function from the hypothesis set. In terms of examples, the Hypothesis set could be linear regression and the learning algorithm could be gradient descent or the hypothesis set could be neural networks and the learning algorithm could be back propagation. Therefore, in order words, hypothesis set is a set of similar functions which have been shown to have good results under specific conditions and the learning algorithm is the algorithm that will do the actual searching." https://medium.com/technology-nineleaps/what-is-machine-learning-8f4f117a9f4e https://medium.com/technology-nineleaps/what-is-machine-lear... I agree with you, I don't find talking about "hypothesis sets" to be at all enlightening. To me it sounds like needless jargon and it is strange to see even this amount of effort expended to use it. Maybe we are missing something here though.

vosper 9y ago

This may be a very naïve question, but how far are we from toolkit that takes a dataset and a scoring function or goal and figures out which is the best technique, and optimizes it too? Sure, it's probably really resource intensive, but can it be done? Maybe with some kind of genetic algorithm selecting techniques and parameters? Soneone must be working on this.

flatfilefan 9y ago
An example of such tool https://github.com/paypal/autosklearn-zeroconf https://github.com/paypal/autosklearn-zeroconf There are many more including genetic algorithms. Some are discussed here: https://www.reddit.com/r/MachineLearning/comments/6efs8u/d_anyone_have_experience_with_automated_ml_tools/ https://www.reddit.com/r/MachineLearning/comments/6efs8u/d_a...
- wakkaflokka 9y ago
  TPOT is also a good automl package that utilizes sklearn in Python. It's also actively under development. (I'm not affiliated, just an occasional user) https://github.com/rhiever/tpot https://github.com/rhiever/tpot
cf 9y ago
The keyword you are looking for is AutoML. http://automl.info/ http://automl.info/
onuralp 9y ago
You might also be interested to check out Automatic Statistician (developed by Z. Ghahramani, J. Tenenbaum and others) - https://www.automaticstatistician.com/index/ https://www.automaticstatistician.com/index/