Is there a difference between the two, or are they different names for the same algorithm?
RandomCutForest (RCF) is an unsupervised method primarily used for anomaly detection, while RandomForest (RF) is a supervised method that can be used for regression or classification.
For RCF, see documentation (here) and notebook example (here)
Related
Does avyone know the reason why the Non-Linear SVM has not been implemented in Apache Spark?
I was reading this page:
https://issues.apache.org/jira/browse/SPARK-4638
Look at the last comment. It says:
"Commenting here b/c of the recent dev list thread: Non-linear kernels for SVMs in Spark would be great to have. The main barriers are:
Kernelized SVM training is hard to distribute. Naive methods require a lot of communication. To get this feature into Spark, we'd need to do proper background research and write up a good design.
Other ML algorithms are arguably more in demand and still need improvements (as of the date of this comment). Tree ensembles are first-and-foremost in my mind."
The question is: Why is the kernelized SVM hard to distribute?
Everybody knows that the non-linear SVMs exhibit better performance than the linear ones.
Currently for face detection I am using svm classifier over HOG feature set.But I need to implement other classifiers over those HOG feature set and compare the results between them .What other classifiers can I use other than svm?
There are plenty of other classification algorithms, a simple logistic regression could be a starting point. You could use logistic regression; implement gaussian process based classifier or random forrests/decision trees and many others with respective pro's and cons.
See e.g. this link for an overview
My question is can we use CNN for feature extraction and then can we use this extracted feature as an input to another classification algorithm like SVM.
Thanks
Yes, this has already been done and well documented in several research papers, like CNN Features off-the-shelf: an Astounding Baseline for Recognition and How transferable are features in deep neural networks?. Both show that using CNN features trained on one dataset, but tested on a different one usually perform very well or beat the state of the art.
In general you can take the features from the layer before the last, normalize them and use them with another classifier.
Another related technique is fine tuning, where after training a network, the last layer is replaced and retrained, but previous layers' weights are kept fixed.
I am trying to implement SVM for multiclass problems in Matlab. I know that there is an inbuilt code for SVM in matlab but I don't know how to use it. Need some help in getting started with Matlab SVM.
SVM classifies into two classes. If you want to create a multiclass SVM, you will have to hack it yourself. You could for instance do AdaBoost with SVMs as your "cheap classifiers", although they are not that cheap to train (contrary to decision trees or even decision stumps).
Speaking of AdaBoost, you'll probably end up using ensemble methods in matlab if you really don't want to program it yourself:
For classification with three or more classes:
'AdaBoostM2'
'LPBoost' (requires an Optimization Toolbox license)
'TotalBoost' (requires an Optimization Toolbox license)
'RUSBoost'
'Subspace'
'Bag'
The ensemble toolbox is really simple and there's a ton of documentation on matlab's help pages. Basically you state your X and Y, the type of learner you want (for instance SVM) and the ensemble method, which is the method you want to use to combine the different weak learners. AdaBoost is one way, but you could also just do Bagging where the majority vote of all your weak learners counts.
So some questions you can answer here or at least ask yourself are: Why do you want to to multiclass SVM? Is it a homework assignment? Do you know how SVM and other machine learning alorithms work? Do you need help picking the right algorithm?
Do you know an example that makes feature selection using a perceptron, maybe an implementation on matlab...
The perceptron is a binary linear classifier, that is, it can classify n-dimensional data that look like this:
but not like this:
into two distinct categories. Just like any other neural network, it first needs to be trained on a training set, and then only it can be used to classify new data points.
The perceptron can therefore be applied to classify any linearly separable dataset. A Matlab implementation is available in the Neural Network toolbox (see the documentation). An excellent toolbox for pattern recognition in general, with excellent classifiers, is PRTools, which is kind of the open source variant of the commercial toolbox PRSD Studio.