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

learn more… | top users | synonyms

1
vote
0answers
49 views

Model selection in offline vs. online learning

I've been trying to learn more about online learning lately (it's absolutely fascinating!), and one theme that I haven't been able to get a good grasp on is how to think about model selection in ...
0
votes
0answers
10 views

Feature Selection Limitations - Machine Learning FPR Test

I am trying to improve accuracy of my logistic regression model by selecting the best features. I did an FPR test and ranked the features based on their F-score. The problem is that selecting the ...
1
vote
1answer
24 views

Why are linear SVMs used with HoG feature descriptors?

Ok, almost all applications I have seen that use HoG features use linear svm as classifier. Can someone explain for me why linear svm are chosen and why they give good performance? Are linear svm ...
0
votes
0answers
37 views

What is the Probability Distribution of NLTK Naive Bayes?

As I know Naïve Bayes has various distributions, as said in Sci-kit learn manual “The different naive Bayes classifiers differ mainly by the assumptions they make regarding the distribution of P(x_i ...
0
votes
0answers
22 views

How to implement data I have to svmtrain() function in MATLAB?

I have to write a script using MATLAB which will classify my data. My data consists of 1051 web pages (rows) and 11000+ words (columns). The first 230 rows are about computer science course (to be ...
0
votes
0answers
15 views

Pseudo-likelihood and RBMs

I need to train a restricted Boltzmann machine to model the joint probability of categorical variables. For this I adapted a Bernoulli RBM to have groups of softmax units in the visible layer. The ...
0
votes
0answers
7 views

maximize mean F1 score in multilabel information retrieval problem

I have a multilabel text classification problem where each observation will have one or more labels associated to it. The metric I want to maximize is mean F1 score. Are there standard ways to ...
0
votes
0answers
25 views

How to determine if a risk indicator is really a risk indicator

Let's say I ask a robot a question: "Why is partial differential equations no fun at all?" Suppose the robot is stupid and categorizes my question inside the class of "cats" so it replies with: ...
0
votes
1answer
39 views

What is the difference between Binary Clasification and Multiclass classification?

Apology for posting almost one question daily. I am trying to learn some aspects of Statistical Machine learning, so every day many questions coming and if I am not finding answer in my offline peer ...
0
votes
0answers
23 views

How to process triaxial accelerometer data for finger taps in real time?

I have a triaxial accelerometer mounted on my index finger and I want to differentiate in real time when it taps on a touch screen vs when the middle finger taps. For this I take the accelerometer ...
0
votes
0answers
25 views

What machine learning tool is best suited for taking time series data as well as descriptive data and making a binomial classification

I have an interesting task of utilizing log data from computer servers in a server farm and predicting if a particular server is likely to fail in the next 24 hours. My data set will be comprised of ...
1
vote
1answer
50 views

Best metric for evaluation of mixture-of-Gaussian clusters on big-data

I have made a new algorithm that is specifically crafted for clustering very large datasets. In order to document it as a research paper, I have to choose one or two internal (no-label) cluster ...
1
vote
1answer
36 views

What does the parameter $\alpha$ do in the Jaccard method for binaryRatingsMatrix in R recommenderlab?

What is the role of the parameter 'alpha' in the recommenderlab R package's use of Jaccard method in the recommender model for ...
2
votes
2answers
78 views

Model Tuning and Model Evaluation in Machine Learning

Despite my readings (on stack 1, 2, or in literature (Cawley, 2010; Japkowicz, 2011)), I don't find a clear procedure for tuning and evaluating a model in a classification task. I want to perform a ...
1
vote
1answer
26 views

What is the training score used for?

This is a pretty basic question. However, I read that the training score isn't useful in machine learning. From what I can tell, the training score only tells us if we have an overfit model. [E.g. if ...
3
votes
2answers
62 views

Is this the definition of over-fitting?

Overfitting is when we have a model which has memorized the training data and does not perform well in real-world cases. Okay, say that I had some training points which look like this: What if the ...
9
votes
4answers
455 views

Hold-out Validation vs K-Fold Validation?

To me, it seems that Hold-out validation is useless. That is, splitting the original dataset into two-parts (training and testing) and using the testing score as a generalization measure, is somewhat ...
1
vote
1answer
23 views

When should a weighted KNN be used (or not)?

By default, machine learning packages turn inverse distance weighting off for KNN. To me, it seems that inverse distance weighting is always a good option. Why would we not want to use IDW with KNN? ...
0
votes
1answer
43 views

How do I detect state change in multivariate time series?

I have a multivariate time series . For each row in the data we have the values of inputs and a label for stability (0 or 1 ) . What are the algorithms that can detect the stability for an unlabelled ...
1
vote
2answers
111 views

How to deal with unbalanced data

I'm doing data analysis with a dataset of 11795 data points (with 88 features). 85% (9973 points) of these data points correspond to data points belonging to class 1, 5% (589 points) belong to class 2 ...
0
votes
2answers
40 views

Best feature selection method for naive Bayes classification

i want to make classification with naive Bayes. I have got about 100 Features. Numerical ones as well as categorical ones. Since i want only the most relevant ones to be included for the ...
3
votes
1answer
56 views

Inputs to k-means are often normalized per-feature. Why not fully whiten the data instead?

We often normalize inputs to the k-means algorithm by 1) subtracting the mean on a per-feature basis and 2) dividing by the standard deviation on a per-feature basis. Some rational behind this is ...
0
votes
0answers
13 views

Correlation between sensors

Background: A home wired with multiple sensors, measuring attributes like temperature, light, motion etc. In addition, a multitude of actuators that can carry out an action like opening a door, ...
42
votes
7answers
6k views

Skills hard to find in machine learners?

It seems that data mining and machine learning became so popular that now almost every CS student knows about classifiers, clustering, statistical NLP ... etc. So it seems that finding data miners is ...
0
votes
0answers
8 views

Regularization in multinomial manifold (k-simplex) space

I have a multi-class classification problem. In order to reduce the feature space, I have mapped my weights ($w$) to a multinomial manifold (or k-simplex) and compute the new weights ($v$) as follows: ...
1
vote
1answer
42 views

Estimation in Naive Bayes

I have a very silly question. In Multinomial Naive Bayes Classifier, which parameter estimation do we use, is it Maximum Likelihood or Maximum A Posteriori? If any one of the esteemed members may ...
1
vote
0answers
5 views

Q-learners with descretised Q-values

I have a Q-learning algorithm with a finite number of states. Each state, however is represented by the usual 32bit Q-value. These Q-values allow to carry information about the previous experience far ...
3
votes
2answers
171 views

What is the justification for unsupervised discretization of continuous variables?

A number of sources suggest that there are many negative consequences of the discretization (categorization) of continuous variables prior to statistical analysis (sample of references [1]-[4] below). ...
0
votes
1answer
29 views

How to compare the significance of two models from two different datasets?

I have two different regression models which I learned from two different data sets. Is there any statistical method which shows the significance of models based on the number of parameters and cross ...
2
votes
3answers
136 views

Naive Bayes: Imbalanced Dataset in Real-time Scenario

I am using scikit-learn Multinomial Naive Bayes classifier for binary text classification (classifier tells me whether the document belongs to the category X or not). I use a balanced dataset to train ...
1
vote
1answer
50 views

Image processing with neural network

I am trying to learn how Neural Network works on image recognition.I am confused that how neural network that how i will give input.My defination is find(track) object in squence of images(particular ...
2
votes
3answers
59 views

Matrix Factorization algorithms for Recommender Systems

I need to learn about Matrix Factorization for recommender systems, so I downloaded this paper https://datajobs.com/data-science-repo/Recommender-Systems-[Netflix].pdf but I found it too shallow. It ...
0
votes
1answer
30 views

Identifying subsets for outlier detection in local outlier factor

I am trying to gain better understanding of the idea of local outliers (as discussed in this pdf) and how the function is implemented. Here are the key passages from the pdf: Local outliers: ...
0
votes
0answers
18 views

What are the best criteria to select the model for Lasso regression?

I have two different formulations of the Lasso regression for the same problem. For each formulation, I selected the best model based on cross validation error. But Now, I want to compare two models ...
1
vote
1answer
31 views

Is there a stats tool for this analysis I run in excel?

I am trying to find a statistical or machine learning tool that replicates this analysis I am doing manually in excel. Each row in my data set is a user. ...
7
votes
2answers
314 views

Why would anyone use KNN for regression?

From what I understand, we can only build a regression function that lies within the interval of the training data. For example (only one of the panels is necessary): How would I predict into the ...
0
votes
1answer
47 views

Regression model for $f(x_1, x_2) = a + b x_1\log x_2$

Which regression algorithm do I need to use to fit the coefficients of $f(x_1, x_2) = a + b x_1\log x_2$? Will linear regression with an independent variable $x_1 \log x_2$ work?
1
vote
1answer
34 views

How do I perform an IDF calculation?

How do I perform an IDF calculation? I am uncertain as to whether IDF should be calculated in per-class level or for the entire document set (that contains multiple classes).
1
vote
0answers
40 views

Does lots of bias==underfitting, while lots of variance==overfitting?

From what I understand, there is a relationship between bias and underfitting; as well as variance and overfitting. Is a 'biased model' another word for an 'underfitted model'? Likewise, is a ...
1
vote
0answers
22 views

Open, multi-class , active learning classfiers

I am trying to classify text documents using a huge corpora. Thats a huge tagset (more than 1000 tags). Corpus will have 1000 samples for each tag. But the tagset is not closed. New tags can be added ...
1
vote
1answer
39 views

Relation between Gaussian mixture models and maximum likelihood?

I need some help understanding the relation between the maximum likelihood and Gaussian mixture models. I have seen that there is a relationship between the expectation maximization algorithms and ...
1
vote
1answer
88 views

What is meant by binary images of handwritten digits?

I can understand what a handwritten digit is, but what is meant by "binary image"? Can someone explain it? I am studying machine learning concepts. In doing so, I've come across many types of inputs. ...
0
votes
1answer
31 views

Question about normalize/scale data before using neuralnet

I have read several threads about the issue on same outputs after people fitting a neural network model with R neuralnet. Posted Solution is to normalize or scale the data before fitting model. Since ...
0
votes
0answers
18 views

Classification More Robust Than Regression

Are neural networks using classification more robust / reliable than using regression to produce a single value? The only reason I would think so is that it would be easier for the network to adjust ...
-1
votes
1answer
38 views

Which machine learning algorithm is the slowest BUT surest?

CrossPost: https://stackoverflow.com/questions/24301743/which-machine-learning-algorithm-is-the-slowest-but-surest?noredirect=1#comment37556042_24301743 Perhaps my perception of time is augmented by ...
0
votes
0answers
18 views

How Sensitive Are Neural Networks?

CrossPost: https://stackoverflow.com/questions/24301472/how-sensitive-are-ff-neural-networks I am aware of pruning, and am not sure if it removes the actual neuron or makes its weight zero, but I am ...
0
votes
1answer
25 views

document similarity with documents using synonyms

I have a bunch of documents where some of the documents are a copy of other documents with their text jumbled up and some of the words replaced by their synonyms. Mentioned below is one such example ...
0
votes
0answers
26 views

Combining CostSensitiveClassifer with MultiClassClassifier [RWeka]

this is my first attempt at posting here. I looked through CrossValidated and found two similar problems without answers (see How to combine WEKA classifiers and Combine MultiClassClassifier and ...
0
votes
1answer
13 views

Solving feature bias issues in Learning to Rank with implicit feedback

I have a learning to rank system where implicit feedback (from user clicks) is used to determine +ve and -ve examples for the training. The problem is that (obviously) the learner sees only the top ...
0
votes
0answers
24 views

Matrix completion approaches for healthcare big data

I am working on a prediction problem that leverage sparse clinical datasets. Missing data rate is in the range of 80%. 1- I am wondering if there is any example of application of matrix completion ...