Methods and principles of building "computer systems that automatically improve with experience."

learn more… | top users | synonyms (1)

0
votes
0answers
40 views

How to compute/run LDA with 3 classes

I couldn't find one example on how to compute LDA with 3 classes (nor what is the algorithm). for example i have the following observations and classes: (each observation in one-dimensional) $ ...
1
vote
0answers
123 views

Performance of algorithms using Jacobian matrix on large data sets

Some ANN learning algorithms like Gauss-Newton, Levenberg-Marquardt require creation of a Jacobian matrix. Having studied this I found out that the Jacobian matrix is huge (unless I misunderstood ...
0
votes
0answers
53 views

Is there “infinite” universal model selection ? and Structural Risk Minimization

I ask these because I come up with an idea : If I have infinite and universal model set, then there must exist model that totally fits my data and no parameter for the model so the complexity ...
0
votes
1answer
305 views

Appropriate non-parametric post-hoc test for baseline comparisons?

I want to evaluate several "classifiers" (machine-learning algorithms) with paired samples. I do not want to compare each algorithms' performance to every other (n x m comparison) but only compare the ...
6
votes
1answer
2k views

what does the numbers in the classification report of sklearn mean?

I have below an example i pulled from sklearn 's sklearn.metrics.classification_report documentation. What i dont understand is why there are f1-score, precision and recall values for each class ...
3
votes
3answers
356 views

Good sources to get papers in machine learning?

I'm a CS master student (will be a PhD in three months or so). Today I was at my superviser's office and he had a friend discussing some ideas for their research. Then they mentioned a paper that I ...
1
vote
1answer
275 views

How to superimpose two MATLAB images rigidly transformed to perform a metrics

I got two MRI images on MATLAB, I need to perform an intensity based registration similarity metrics in order to get a registration. The problem is that, since I got some rigid transformation on one ...
1
vote
1answer
260 views

Weight Decay in Neural Neural Networks Weight Update and Convergence

I have a neural network (That I created using java) for a class assignment that is working when I do not use any weight decay value, but when I use a value greater than or equal to .001, my accuracy ...
2
votes
1answer
321 views

Logistic-Regression: Prior correction at test time

Using sklean.linear_model.LogisticRegression for a binary classification problem. My classes are unbalanced. The positive class comprises about 20% of the training set. When fitting the model I use: ...
0
votes
1answer
277 views

Crossover and Mutation in Genetic Algorithm

I am studying how GA works. It is known that GA obtains the optimal solution by iteration through the the process of reproduction, crossover, and mutation. When crossover and mutation, the ...
5
votes
3answers
973 views

Understanding Gaussian Basis function parameters to be used in linear regression

I'd like to apply the Gaussian basis function into a linear regression implementation. Unfortunately I'm having a hard time understanding a couple parameters in the basis function. Specifically mu ...
1
vote
0answers
130 views

Machine Learning Models For Real Time Sales Data

I am working on a predictive analytics problem related to Sales where based on interaction with a prospect we try to predict whether the deal will close or not. In sales, the data updates whenever a ...
1
vote
1answer
71 views

A simple question on CLT in possible connection with Berry-Esseen thm

I am curious about the contents while I read a note on machine learning. It could be obvious. So, please let me know if I am missing some fundamental things. $X_1,X_2,...,X_n$ are from an i.i.d. ...
0
votes
0answers
123 views

Neural Networks sigmoid activation with bias updates

I am trying to figure out if I am creating an artificial neural network using the sigmoid activation function and using bias correctly. I want one bias node to input to all hidden nodes with static ...
3
votes
2answers
2k views

Pattern recognition with time series analysis

I'm looking for some good pointers to pattern recognition with time series. Possibly something practical that can be easily understood. As a toy example, think about collecting data from an ...
0
votes
1answer
134 views

Use of infinity norm instead of SSE for machine learning accuracy?

Are there any examples or arguments in favor of using an infinity norm (or equivalent) over sum of squared errors or root mean squared error for evaluating machine learning algorithms?
1
vote
1answer
93 views

Strategy for Analyzing Data

I have been learning about Machine Learning (via Udacity) and Statistics (via Coursera) the past few months and trying to figure out a good way to combine them for a general approach to explaining ...
0
votes
0answers
158 views

What is the next step after acquiring the parameters(means, covar, priors) from GMM via EM

I am comparing the results achieved from clustering via K-means and GMM. For comparison I have accumulated a dataset of images. The training set consists of 359 images. I used SIFT to extract the ...
1
vote
0answers
52 views

Dichotomizing Continuous Variables in Regression: Good or Bad? [duplicate]

I believe Dichotomizing(also called bucketing/binning) of continuous variable is not always a good idea. My colleague while building regression model always bins continuous variables and only keep ...
0
votes
0answers
146 views

Intuition on One Class Support Vector Machines

I am having trouble understanding how one-class SVMs work. They were introduced in a paper by Scholkopf and others (and can be found here). One-class SVMs perform "novelty detection", where a point ...
0
votes
0answers
268 views

What is a convolutional neural network

I have been studying neural networks and I recently found out about deep learning and convolutional neural networks. Can someone give me a newbie introduction to convolutional neural networks, what ...
1
vote
1answer
81 views

Affinity propagation comments

I was looking into affinity propagation for my similarity matrix problem and thought it would fit well. However, browsing literature I found this comment that basically breaks both legs of affinity ...
1
vote
2answers
134 views

How to partition a training-set when I have a big class imbalance?

In my actual data class A has 90%, class B has 9% and class C has 1% (numbers are made up for sake of simplicity). Now I want to prepare a training set for my classifier (I plan to use Vowpal Wabbit). ...
0
votes
2answers
41 views

What's the best algorithm type for low-dimensional grouping

I'm looking for some advice on directions to head in a project I'm working on. Basically what I want to do is identify general (of varying size) groups in a 2-D grid of points belonging to one of ...
2
votes
2answers
227 views

Understanding differences between large and small dimensional data when implementing algorithms

I'm working on a local outlier factor implementation based on the wikipedia entry : http://en.wikipedia.org/wiki/Local_outlier_factor This article seems to explain it in just two dimensional data. ...
0
votes
1answer
184 views

Question of unary term (data term) of the graph cut method

I am trying to apply graph cut method for my segmentation task. I found some example codes at Graph_Cut_Demo. Part of the codes are showing below ...
1
vote
0answers
862 views

predicting time series with support vector machine using R

I am planning to do time series prediction using support vector Machine. I could not find any materials about time series application of support vector machines using R or Mat-lab. Similar question ...
3
votes
1answer
855 views

In neural networks, how to tell the feature which contributes the most to the output value?

I have a neural network which uses input features $N$, some input layers $L$ which predict a continuous variable $X$. Can we say which features or combination of 2 features of the initial $N$ features ...
1
vote
2answers
428 views

Varying LIBSVM predictions based on test series labels

So I have a pretty well testing SVC train series which puts me into the mid 80 percentile without outrageous C/g values. My current C value is 2.0 and gamma is 0.5. Good numbers across the range ...
0
votes
1answer
1k views

How to prepare my data for SVM classifier in matlab

I am new to SVM and Matlab. I would like to have an example how to prepare my data to be as input to the SVM classifer (using libsvm) let us assume that i have a group of words first i have ...
0
votes
1answer
86 views

Can adaboost choose the same variable for multiple splits for a given tree?

Can adaboost choose the same variable for multiple splits for a given tree? The model was given 100 + variable to choose from and it did choose them for the other trees in the ensemble. I am using ...
2
votes
1answer
65 views

Overfitting with low test error

I have an ensemble of models that yields overfitting by some models by looking at the differentce between training and test error and also overfits in the combined ensemble of models providing a very ...
1
vote
1answer
102 views

Requirements for a valid neural network activation function?

What rules define a valid neural network activation function, excluding biological plausibility? What set of principles do softmax, rectified linear units, hyperbolic tangent, sigmoid, etc. follow? ...
2
votes
1answer
121 views

How to find all optima in an optimization problem?

I have an optimization problem where several optima can exist at different input values, and I need to find as many as possible. As an example consider the cross-in-tray function, which has four ...
3
votes
0answers
215 views

What enforces features diversity in RBM?

I'm working on an implementation of a Restricted Boltzman Machine (RBM). I made some tests on the MNIST dataset trying to learn a representation of the digit 2. My inputs are binary images. My aim is ...
0
votes
1answer
164 views

Suggestions needed about classifier fusion

I'm working on a classification problem which involves two classifier to observe a single event. I'm providing a high level description of the problem without going into the technical details (the ...
0
votes
1answer
86 views

Machine learning - curve categorisation

I have curves of the following structure (it is the blue one I am interested in) These curves reflect the volume of blood (actually gamma ray counts) in the left ventricle as a function of time ...
0
votes
1answer
51 views

Statistical Commute Analysis in Java

I have a rather large commute every day - it ranges between about an hour and about an hour and half of driving. I have been tracking my driving times, and want to continue to do so. I am capturing ...
3
votes
3answers
148 views

Optimal classification model for translating words

I have the following problem: I have a set of English words which I want to translate to Dutch. Of each words I mined a set of possible translations. For example, for the word "Eighteen" I obtained ...
1
vote
1answer
374 views

Algorithm to find subsets with high correlation

I have a reasonably large dataset (d) with predictor variables x1...xn and a target variable y. I can use recursive partitioning (such as CART or rpart in R) to find subsets of d with a high (or low) ...
2
votes
1answer
183 views

Working with few data examples

I have been asked often in some interview, that how we should proceed when we have less data examples(say 50 or 100). What considerations needs to be made while choosing any algorithm. few points ...
0
votes
1answer
40 views

How to improve the performance of K-nn algorithm in R?

I am having a digit recognizer data set which has column names as label, pixel0, pixel1...pixel783. pixel values vary from 0 to 255 indicating the lightness or darkness of that pixel, with higher ...
2
votes
0answers
310 views

How to split a decision tree when information gains of all attributes are zero?

The textbook tells us that we should choose an attribute with the maximum information gain to split a decision tree. My question is what if all information gains are zero? Should we stop splitting or ...
0
votes
0answers
47 views

Best classifiers for large data sets?

I'm working on a data set that contains electricity consumption data. There will be 2-3 features used. I'm not sure if that is all of the features to be used. Also, it will be a really large data set. ...
1
vote
0answers
17 views

How do i generate variables that are relevant only for some classes? [closed]

I want to generate data for classification. I've generated data with 10 variables with two are relevant for all classes and 8 noise. now, I want to generate variables that are relevant just for some ...
0
votes
0answers
160 views

How to represent outliers for multi dimensional data (local outlier factor)

Below graph taken from http://en.wikipedia.org/wiki/Local_outlier_factor displays "LOF scores : LOF image : This is great for two dimensional data but what about data > than two dimensions. How ...
1
vote
2answers
138 views

Newbie: Decision Tree in R with Two classes(Yes/No) where one class (No) is much larger than other class (Yes)

I am trying to make a decision tree using 4 features (A,B,C,D) to predict an out come for two classes E(Yes, NO). The problem is that the number of observations in my dataset that belong to one class ...
2
votes
1answer
38 views

What kind of functions can have non whole degrees?

Thanks for the help in advance. I am reading a technical report on a regression algorithm that reports a pair of functions as having a total degree of freedom of 5.4. I believe that both of these ...
1
vote
1answer
76 views

Selecting most realistic C and g params after gridsearch

I just ran an extended SVC gridsearch in libsvm on about 9000 multi-dimensional vectors representing a time series. Here are the highest scoring results: ...
1
vote
1answer
171 views

How to evaluate a clustering/unsupervised learning problem with massive amounts of data, with labels only for a small fraction of points

I'm wondering if anybody can point me to work on the evaluation of unsupervised learning where there are a very large (say hundreds of millions) number of points and manual labelling can only ever be ...