I'm working on a machine learning problem which requires me to use multiple support vector machines. It works relatively well; however, the problem is that the number of support vector machines for each SVM tends to be large (~2,000), and the number of input features is on the order of 50,000. I need about a 100 SVMs.
Running this on my laptop quickly exhausts all of the available memory; I think this is because svmtrain creates an SVM which saves (i.e. has another copy) of all the support vectors. Since I'm keeping the original training data around, I was wondering if there was a way to instruct it to save indexes to these support vectors instead, which would use much less memory? Or another way to reduce the amount of memory it needs per SVM?
svmtrain creates SVMStruct object, which stores the indices of support vectors in SupportVectorIndices field. So simply store this variable's values in some container and free the rest of the model.
Accodring to documentation
SupportVectorIndices — Vector of indices that specify the rows in Training, the training data, that were selected as support vectors after the data was normalized, according to the AutoScale argument.
Related
I am working on a Information Retrieval model called DPR which is a basically a neural network (2 BERTs) that ranks document, given a query. Currently, This model is trained in binary manners (documents are whether related or not related) and uses Negative Log Likelihood (NLL) loss. I want to change this binary behavior and create a model that can handle graded relevance (like 3 grades: relevant, somehow relevant, not relevant). I have to change the loss function because currently, I can only assign 1 positive target for each query (DPR uses pytorch NLLLoss) and this is not what I need.
I was wondering if I could use a evaluation metric like NDCG (Normalized Discounted Cumulative Gain) to calculate the loss. I mean, the whole point of a loss function is to tell how off our prediction is and NDCG is doing the same.
So, can I use such metrics in place of loss function with some modifications? In case of NDCG, I think something like subtracting the result from 1 (1 - NDCG_score) might be a good loss function. Is that true?
With best regards, Ali.
Yes, this is possible. You would want to apply a listwise learning to rank approach instead of the more standard pairwise loss function.
In pairwise loss, the network is provided with example pairs (rel, non-rel) and the ground-truth label is a binary one (say 1 if the first among the pair is relevant, and 0 otherwise).
In the listwise learning approach, however, during training you would provide a list instead of a pair and the ground-truth value (still a binary) would indicate if this permutation is indeed the optimal one, e.g. the one which maximizes nDCG. In a listwise approach, the ranking objective is thus transformed into a classification of the permutations.
For more details, refer to this paper.
Obviously, the network instead of taking features as input may take BERT vectors of queries and the documents within a list, similar to ColBERT. Unlike ColBERT, where you feed in vectors from 2 docs (pairwise training), for listwise training u need to feed in vectors from say 5 documents.
I have a matrix X, size 40-by-60000
while writing the SVM, I need to form a linear kernel: K = X'*X
And of course I would get an error
Requested 60000x60000 (26.8GB) array exceeds maximum array size preference.
How is it usually done? The data set is Mnist, so someone must have done this before. In this case rank(K) <= 40, I need a way to store K and later pass it to quadprog.
How is it usually done?
Usually kernel matrices for big datasets are not precomputed. Since optimisation methods used (like SMO or gradient descent) do only need access to a subset of samples in each iteration, you simply need a data structure which is a lazy kernel matrix, in other words - each time an optimiser requests K[i,j] you literally compute K(xi,xj) then. Often, there are also caching mechanisms to make sure that often requested kernel values are already prepared etc.
If you're willing to commit to a linear kernel (or any other kernel whose corresponding feature transformation is easily computed) you can avoid allocating O(N^2) memory by using a primal optimization method, which does not construct the full kernel matrix K.
Primal methods represent the model using a weighted sum of the training samples' features, and so will only take O(NxD) memory, where N and D are the number of training samples and their feature dimension.
You could also use liblinear (if you resolve the C++ issues).
Note this comment from their website: "Without using kernels, one can quickly train a much larger set via a linear classifier."
This problem occurs due to the large size of your data set, thus it exceeds the amount of RAM available in your system. In 64-bit systems data processing performs better than in 32-bit, so you'll want to check which of the two your system is.
I have a 40X3249 noisy dataset and 40X1 resultset. I want to perform simple sequential feature selection on it, in Matlab. Matlab example is complicated and I can't follow it. Even a few examples on SoF didn't help. I want to use decision tree as classifier to perform feature selection. Can someone please explain in simple terms.
Also is it a problem that my dataset has very low number of observations compared to the number of features?
I am following this example: Sequential feature selection Matlab and I am getting error like this:
The pooled covariance matrix of TRAINING must be positive definite.
I've explained the error message you're getting in answers to your previous questions.
In general, it is a problem that you have many more variables than samples. This will prevent you using some techniques, such as the discriminant analysis you were attempting, but it's a problem anyway. The fact is that if you have that high a ratio of variables to samples, it is very likely that some combination of variables would perfectly classify your dataset even if they were all random numbers. That's true if you build a single decision tree model, and even more true if you are using a feature selection method to explicitly search through combinations of variables.
I would suggest you try some sort of dimensionality reduction method. If all of your variables are continuous, you could try PCA as suggested by #user1207217. Alternatively you could use a latent variable method for model-building, such as PLS (plsregress in MATLAB).
If you're still intent on using sequential feature selection with a decision tree on this dataset, then you should be able to modify the example in the question you linked to, replacing the call to classify with one to classregtree.
This error comes from the use of the classify function in that question, which is performing LDA. This error occurs when the data is rank deficient (or in other words, some features are almost exactly correlated). In order to overcome this, you should project the data down to a lower dimensional subspace. Principal component analysis can do this for you. See here for more details on how to use pca function within statistics toolbox of Matlab.
[basis, scores, ~] = pca(X); % Find the basis functions and their weighting, X is row vectors
indices = find(scores > eps(2*max(scores))); % This is to find irrelevant components up to machine precision of the biggest component .. with a litte extra tolerance (2x)
new_basis = basis(:, indices); % This gets us the relevant components, which are stored in variable "basis" as column vectors
X_new = X*new_basis; % inner products between the new basis functions spanning some subspace of the original, and the original feature vectors
This should get you automatic projections down into a relevant subspace. Note that your features won't have the same meaning as before, because they will be weighted combinations of the old features.
Extra note: If you don't want to change your feature representation, then instead of classify, you need to use something which works with rank deficient data. You could roll your own version of penalised discriminant analysis (which is quite simple), use support vector machines, or other classification functions which don't break with correlated features as LDA does (by virtue of requiring matrix inversion of the covariance estimate).
EDIT: P.S I haven't tested this, because I have rolled my own version of PCA in Matlab.
I am studying Support Vector Machines (SVM) by reading a lot of material. However, it seems that most of it focuses on how to classify the input 2D data by mapping it using several kernels such as linear, polynomial, RBF / Gaussian, etc.
My first question is, can SVM handle high-dimensional (n-D) input data?
According to what I found, the answer is YES!
If my understanding is correct, n-D input data will be
constructed in Hilbert hyperspace, then those data will be
simplified by using some approaches (such as PCA ?) to combine it together / project it back to 2D plane, so that
the kernel methods can map it into an appropriate shape such a line or curve can separate it into distinguish groups.
It means most of the guides / tutorials focus on step (3). But some toolboxes I've checked cannot plot if the input data greater than 2D. How can the data after be projected to 2D?
If there is no projection of data, how can they classify it?
My second question is: is my understanding correct?
My first question is, does SVM can handle high-dimensional (n-D) input data?
Yes. I have dealt with data where n > 2500 when using LIBSVM software: http://www.csie.ntu.edu.tw/~cjlin/libsvm/. I used linear and RBF kernels.
My second question is, does it correct my understanding?
I'm not entirely sure on what you mean here, so I'll try to comment on what you said most recently. I believe your intuition is generally correct. Data is "constructed" in some n-dimensional space, and a hyperplane of dimension n-1 is used to classify the data into two groups. However, by using kernel methods, it's possible to generate this information using linear methods and not consume all the memory of your computer.
I'm not sure if you've seen this already, but if you haven't, you may be interested in some of the information in this paper: http://pyml.sourceforge.net/doc/howto.pdf. I've copied and pasted a part of the text that may appeal to your thoughts:
A kernel method is an algorithm that depends on the data only through dot-products. When this is the case, the dot product can be replaced by a kernel function which computes a dot product in some possibly high dimensional feature space. This has two advantages: First, the ability to generate non-linear decision boundaries using methods designed for linear classifiers. Second, the use of kernel functions allows the user to apply a classifier to data that have no obvious fixed-dimensional vector space representation. The prime example of such data in bioinformatics are sequence, either DNA or protein, and protein structure.
It would also help if you could explain what "guides" you are referring to. I don't think I've ever had to project data on a 2-D plane before, and it doesn't make sense to do so anyway for data with a ridiculous amount of dimensions (or "features" as it is called in LIBSVM). Using selected kernel methods should be enough to classify such data.
I have a dataset of n data, where each data is represented by a set of extracted features. Generally, the clustering algorithms need that all input data have the same dimensions (the same number of features), that is, the input data X is a n*d matrix of n data points each of which has d features.
In my case, I've previously extracted some features from my data but the number of extracted features for each data is most likely to be different (I mean, I have a dataset X where data points have not the same number of features).
Is there any way to adapt them, in order to cluster them using some common clustering algorithms requiring data to be of the same dimensions.
Thanks
Sounds like the problem you have is that it's a 'sparse' data set. There are generally two options.
Reduce the dimensionality of the input data set using multi-dimensional scaling techniques. For example Sparse SVD (e.g. Lanczos algorithm) or sparse PCA. Then apply traditional clustering on the dense lower dimensional outputs.
Directly apply a sparse clustering algorithm, such as sparse k-mean. Note you can probably find a PDF of this paper if you look hard enough online (try scholar.google.com).
[Updated after problem clarification]
In the problem, a handwritten word is analyzed visually for connected components (lines). For each component, a fixed number of multi-dimensional features is extracted. We need to cluster the words, each of which may have one or more connected components.
Suggested solution:
Classify the connected components first, into 1000(*) unique component classifications. Then classify the words against the classified components they contain (a sparse problem described above).
*Note, the exact number of component classifications you choose doesn't really matter as long as it's high enough as the MDS analysis will reduce them to the essential 'orthogonal' classifications.
There are also clustering algorithms such as DBSCAN that in fact do not care about your data. All this algorithm needs is a distance function. So if you can specify a distance function for your features, then you can use DBSCAN (or OPTICS, which is an extension of DBSCAN, that doesn't need the epsilon parameter).
So the key question here is how you want to compare your features. This doesn't have much to do with clustering, and is highly domain dependant. If your features are e.g. word occurrences, Cosine distance is a good choice (using 0s for non-present features). But if you e.g. have a set of SIFT keypoints extracted from a picture, there is no obvious way to relate the different features with each other efficiently, as there is no order to the features (so one could compare the first keypoint with the first keypoint etc.) A possible approach here is to derive another - uniform - set of features. Typically, bag of words features are used for such a situation. For images, this is also known as visual words. Essentially, you first cluster the sub-features to obtain a limited vocabulary. Then you can assign each of the original objects a "text" composed of these "words" and use a distance function such as cosine distance on them.
I see two options here:
Restrict yourself to those features for which all your data-points have a value.
See if you can generate sensible default values for missing features.
However, if possible, you should probably resample all your data-points, so that they all have values for all features.