Looks like K-nearest neighbor does pretty well.
Try the "Island inside an island" test (put a blue cluster inside an orange island). Only k-means and SVM dealt with it satisfactorily.
Shouldn't a neural net with sufficient unit do that too?
Visualizing popular machine learning algorithms
21–30 of 39 posts
Re: Visualizing popular machine learning algorithms
#22Sounds interesting, but I can't see the results with Firefox 38.2.1.
Try this one: https://jsfiddle.net/752pqyvp/embedded/result
It's because of the browser blocking mixed content: The JS libraries are being loaded over HTTP but the JSFiddle is over HTTPS.
The version above loads the libraries over HTTPS via cdnjs.com
Re: Visualizing popular machine learning algorithms
#23Re: Visualizing popular machine learning algorithms
#24No sources on GitHub, but why? Nerdy.js looks like interesting component, but I failed to find any relevant information about it.
Re: Visualizing popular machine learning algorithms
#25Looks like K-nearest neighbor does pretty well.
Yes, k-nn is theoretically the one of the best ML algorithms in the sense that it will find the closest items in the training set. For classification or finding similar looking items it is great. However, it has pretty poor running times for evaluation of unseen data ( http://nlp.stanford.edu/IR-book/html/htmledition/time-comple... ). This is contrary to something like neural networks, which take a while to train, bu…
I wouldn't call it theoretically the best. It is affected by outliers and doesn't make any generalization at training time. This latter point raises the questions whether it deserves the name learning. I would say linear models are typically a better learning algorithm; I wouldn't know what to call "the best" algorithm, but it might be deep learning nowadays.
Re: Visualizing popular machine learning algorithms
#26There is also MLDemos [1] which is open source.
Re: Visualizing popular machine learning algorithms
#27any visualization of these algorithms in 2 dimensions (with cubic feature expansion!) is completely misleading if you intend to work on any real problem with many dimensions. Also, for those asking for execution times, these would be horribly misleading as well.
+1!
Are you aware of reasonable high dimensional "visualizations". It cant' be accurate of course. But catpuring essential features would be nice.
E.g. here is a 4d cube: https://commons.wikimedia.org/wiki/File:8-cell.gif
Re: Visualizing popular machine learning algorithms
#28Earlier quoted context omitted.
Try the "Island inside an island" test (put a blue cluster inside an orange island). Only k-means and SVM dealt with it satisfactorily.
Shouldn't a neural net with sufficient unit do that too?
Yes, if you increase the hidden layer from 5 to 10 nodes:
Re: Visualizing popular machine learning algorithms
#29Looks like K-nearest neighbor does pretty well.
These visualisations are great but misleading regarding the performance of these classifiers. In practice you don't have a lot of data in a small number of dimensions (2 in this case). You have a little bit of data in zillions of dimensions. Think of classifying a 100x100 pixel image: that's 3x100x100=30000 dimensional data. You may not even have one training sample per class per dimension. Generalizing from comparatively little data to a very high dimensional space is the true difficulty of machine learning. Unfortunately you can't easily visualize that.
Re: Visualizing popular machine learning algorithms
#30Can someone please explain this?