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

learn more… | top users | synonyms

0
votes
0answers
28 views

Validation accuracy larger than training accuracy

I was performing an experiment but got a higher validation accuracy than training accuracy. I've got a 39 mice data and performed leave one out cross-validation. The validation accuracy was 100%. But ...
2
votes
2answers
32 views

Accuracy of random prediction with non equal distribution

Assume that I want to predict the value of a variable that has three different states: a, b, and c. The chance that these variables have the 3 states is not equally distributed. Out of 10 trials, the ...
2
votes
2answers
54 views

Which property of count data make mean-variance dependency?

I have read about the fact that, there is dependency of variance on mean of count data.In most of cases they do variance stabilization transfomration as preprocessing step of data modeling. I wonder, ...
0
votes
1answer
30 views

Multi-class Confusion Matrix to Binary confusion matrix

i know the main concepts of data/text mining but i used them mainly in binary classification problems (just two classes). i am now dealing with a problem with 8 classes and i am atruggling how to ...
0
votes
0answers
22 views

How to deal with a feature relating to what type of expert labelled the data that becomes unavailable at point of classification?

Essentially I have a data set, that has a feature vector, and label indicating whether it is spam or non-spam. To get the labels for this data, 2 distinct types of expert were used each using ...
0
votes
1answer
48 views

why use Supervised vs Unsupervised given the class label?

Hi I have data set with a set of variables and known class labels. I am trying to compare why a supervised approach will work theoretically better compared to a unsupervised approach for ...
1
vote
0answers
67 views

MHT: Pre-Selecting statistical tests without Bias

Summary: The last formula boxed in red (which is a modified log likelihood from logistic regression) is a special non-differentiable loss function that is adapted to contain a Bonferroni correction in ...
0
votes
0answers
19 views

auto-steering using neural networks

I was hoping if anyone could point me in right direction, I want to implement a neural network that could steer an autonomous car, I have implemented basic classification problems before using single ...
2
votes
4answers
87 views

Interpreting conflicting results from Random Forest & Logistic Regression?

I am using SKLearn and Statsmodel in python to build a RF and Logistic Regression, respectively. I have a feature that the RF indicates is important (feature importance of 0.202, closely behind #1 ...
1
vote
1answer
20 views

How should the precision/recall be calculated for classes in datasets with NO true class instances?

I have built a classification model to recognise a class and I have evaluated it on several datasets. The problem is that some of these datasets do not have any true instance of the class in question, ...
0
votes
0answers
3 views

what is the meaning of the Samples in NER?

I would like to know in NER (Named Entity Recognition ) problem , which concept should be considered as samples? each token as a sample? or each sentence ? or each Named Entity should be considered ...
0
votes
0answers
16 views

What is the difference between Contrastive divergence k and persistent contrastive divergence algorithm?

As per my understanding Contrastive divergence k is obtaining v(k) after k steps of gibbs chain. Persistent contrastive divergence is obtaining v(k) independent of v(0). I am quite confused with the ...
0
votes
0answers
28 views

Which statistics to use in order to understand a dataset?

So I have a dataset that I will use to train a bunch of classifiers. I need to do that for my thesis. However I'm not sure which statistics are good to use to better understand the dataset and the ...
0
votes
2answers
68 views

Using decision trees to make a binary decision

I have a button that I can press or not press, a binary target that I would like to be 1 as often as possible, and a bunch of features. I also have a bunch of (feature, button choice, target) data, in ...
2
votes
0answers
44 views

Unbalanced dataset - ROC curve to compare classifiers?

I use the machine learning software WEKA for data mining on biological data. I would describe my dataset as unbalanced: It comprises around 2000 instances, ...
1
vote
3answers
72 views

Is it necessary to scale the target value in addition to scaling features for regression analysis?

I'm building regression models. As a preprocessing step, I scale my feature values to have mean 0 and standard deviation 1. Is it necessary to normalize the target values also?
0
votes
0answers
30 views

What is task-loss function?

I looked into "Multi-Output Learning for Camera Relocalization" research and faced with the following part (2.2 The Direct Regression Approach): Given a set of RGB-D frames with known camera poses ...
0
votes
1answer
55 views

How to determine which variable or combination of the variables are affecting to the predictor variable?

I have one dependent variable name as "win ration" of the deal contested and more than 30 independent variables, all are categorical variable name as role of the customer, geo, region, and 27 ...
3
votes
4answers
622 views

Solving a practical machine learning problem

I am currently doing my Phd in computational biology at Stanford. I get the data I need to answer the questions I am interested in. The data sets are sometimes "large" and these large problems take ...
2
votes
2answers
93 views

Reproduce linear discriminant analysis projection plot

I'm struggling with projection points in linear discriminant analysis (LDA). Many books on multivariate statistical methods illustrate the idea of the LDA with the figure below. The problem ...
1
vote
0answers
30 views

Neural Networks and Picture Recognition

I have spent a bunch of time looking at this series of videos (Neural Network Tutorial), by Ryan Harris: https://www.youtube.com/watch?v=Q_5B3GuWPCc&index=41&list=PL29C61214F2146796 I am ...
2
votes
1answer
24 views

How to isolate impact of event in a product's lifecycle?

I'm trying to figure out how a single event affects sales numbers of a song. For example, see what the effect of being featured in iTunes store compared to songs with comparable previous download ...
2
votes
2answers
77 views

Weighting words based on position in text

I'm currently working on semantic analysis and had a question about text organization and structure. Are there any algorithms, or statistical / machine-learning models that weight the importance of a ...
0
votes
0answers
32 views

Obtaining sequence of lambda values for training glmnet model via `caret`

I have multiple models that I'm training using train in the caret package, all while using the same cross validation folds to ...
0
votes
0answers
23 views

Does it work better to subtract mean of data in logistic regression?

I am using logistic regression to predict $X \rightarrow Y \in \{0, 1\}$ based on the featurization $\phi(X)$. The training objective function is \begin{equation} \mathcal{L} = ...
0
votes
1answer
32 views

Why, when I scale my data set, glmnet gives error?

I'm using glmnet for building the regression models. My data are already log-transformed. when I scale my data set (zero mean, and SD=1), I get the following error: ...
1
vote
0answers
19 views

How to give an input when you are using Machine Learning method in R

I am new to R and machine learning algorithms. I have basic knowledge of different machine learning algorithms. I have four years of daily sales data.I am trying to predict sales using Support Vector ...
0
votes
0answers
15 views

How could I generate an “explanation” for each prediction in a classification ?

I have a classification problem. My classes are 0 and 1. The dataset is a bit big, the training is done on 7 million lines and 100 + variables so I choose to use scikit learn and the logistic ...
3
votes
1answer
75 views

Number of Latent Factors in Non-negative Matrix Factorization

Given a matrix $V^{m \times n}$, Non-negative Matrix Factorization (NMF) finds two matrices $W^{m \times k}$ and $H^{k \times n}$ to represent the decomposed matrix as: $V = WH$ Are there common ...
0
votes
0answers
6 views

In LibSVM, svm-scale gives data that is all 1 and -1 [migrated]

As is described in the title, when I try to use svm-scale to scale my regression data into [-1, 1], the scaled data is all 1 or -1. I've confirmed that the original data itself has no problem. I'm on ...
-2
votes
0answers
15 views

What are the applications of multi-agent approach for large scale data mining

I understand that some large scale data mining can be done via grid computing. So I am wondering if there is any advantage of employing multi-agent approach in this area. Any references or resources? ...
0
votes
0answers
22 views

Prediction using Support Vector (SV) method in R

I came to know that using SVM method we can predict the future value more accurately than other normal methods (like ARIMA). My question is how do we give the future index value (let's say 101 when we ...
0
votes
0answers
17 views

GBM, it's overfitting/multicollinearity problem and parameter setting up

I recently came across a predicting problem (0-1 outcome, with more than 80 variables), I decided to use GBM (Gradient Boosting Machine by Friedman)to handle this job. I let the GBM use only 70% of ...
0
votes
0answers
15 views

Removing labelling noise

I have a big data set with unlabelled observations (several million) and about 20 thousand properly labelled ones. There are only two classes and all correctly labelled samples belong to the same ...
0
votes
1answer
26 views

mixing binary and real-valued features with SGD

I'm going to be using a logistic regression model and using SGD to determine the feature weights. Is it OK for me to use a mix of binary and real features, without doing anything like scaling or ...
0
votes
1answer
13 views

Feature selection with a binary dependent variable

Given we have a binary dependent variable and 100s of features and ~50k observations, is there a generally accepted way to trim the features via some type of machine learning concept? I was trying a ...
0
votes
0answers
11 views

Restricted Boltzman Machine Non-Hidden Layer Approach

An RBM is defined by the joint probability distribution $$p({\bf x},{\bf h})=\exp(-E({\bf x},{\bf h}))/Z$$ where $$E({\bf x},{\bf h})=-{\bf h}^TW{\bf x} - {\bf c}^T{\bf x} - {\bf b}^T{\bf h}$$ ...
1
vote
0answers
16 views

Learned production test

To validate the acoustic performance of a product, we are using hand-engineered features and thresholds. Everytime a new hardware problem arises we have to at least tweak a parameter and at worst add ...
9
votes
2answers
628 views

What does the name “Logistic Regression” mean?

I am checking an implementation of Logistic Regression from here. After I reading that article, it seems the important part is the find the best coefficients to determine the sigmoid function. So I ...
0
votes
1answer
30 views

Calculating the information gain on the features with python

I'm looking for a python library that computes the information gain for the features given a training matrix. Are you aware of any?
0
votes
0answers
22 views

semisupervised classification training on all or part of unlabeled data

I have 3 sets of data. A positively labeled dataset. An unlabeled dataset that has for sure positive (around 75%) and negative data. An unlabeled dataset that has for sure positive data and maybe ...
2
votes
2answers
79 views

How to build a prediction model for exam score based on previous scores

I am trying to construct a formula, which will take student's previous exam results (for ex: SAT) taken at particular dates and predict his future test result. One X is previous test result 1; ...
1
vote
2answers
108 views

Multivariate Bayesian formula

I got there example graphs bishop's PRML (8.2.1) 1. a <- c -> b $$ p(a,b,c) = p(a|c)p(b|c)p(c) --(1)\\ p(a,b) = \sum_c p(a|c)p(b|c)p(c) --(2) $$ Q1: Can I use a new graph to represent the ...
1
vote
1answer
23 views

Pairwise compatibility metric

I work at a company that sells clothes, and I've had good results with using cosine similarity to determine which products are "similar" to each other simply based on who owns them. I wanted to take ...
0
votes
0answers
23 views

how to measure the transform ability of a query on Internet Advertising

I have a task which aims to measure the transform ablity of a keyword , the transform ablity can be understood as the probability of a query on the search engine brings an order to the customers.For ...
0
votes
1answer
31 views

How to understand the Gaussian Progress Regression?

I know Gaussian Progress Regression is completed determined by its mean and covariance functions. For given training and test data, I can compute the covariance matrix correctly like a machine, and I ...
0
votes
0answers
9 views

Calculating the Similarity of Survey Responses

I was wondering if anyone had experimented with different functions for calculating the similarity of two sets of survey responses. I am going to be plugging it into a hierarchical clustering algo and ...
2
votes
1answer
42 views

Model for probability of song reaching top 10 ranking, over time?

I'm trying to model the probability of a song reaching Billboards top 10 over time. My data has the columns "Day since release", "If reached top 10". For example, [12,1] means the song hit top 10 on ...
0
votes
0answers
16 views

Confusing related to kernel method of Regression

In the Support Vector Machine (SVM), non-linear data are mapped to a higher-dimensional feature space to be separated linearly, a kernel is used to compute the inner product in the lower dimensional ...
0
votes
3answers
93 views

Is Maximum Likelihood Estimation (MLE) a parametric approach?

There are two main probabilistic approaches to novelty detection: parametric and non-parametric. The non-parametric approach assumes that the distribution or density function is derived from the ...