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

learn more… | top users | synonyms (1)

3
votes
0answers
43 views

What is the intuition behind a Long Short Term Memory (LSTM) recurrent neural network?

The idea behind Recurrent Neural Network (RNN) is clear to me. I understand it in the following way. We have a sequence of observations ($\vec o_1, \vec o_2, \dots, \vec o_n$) (or, in other words, ...
1
vote
0answers
26 views

Large sample with many little groups of dependent observations

I work with traffic crash data and my sample consists of about 165,000 injured people distributed over roughly 107,000 crashes. The prevalent approach in traffic crash analysis is to look at every ...
0
votes
0answers
25 views

No change in accuracy big vs small training set size ConvNet

I am doing some small experiments with image classification in Caffe using the AlexNet architecture. I use a dataset of 50 classes with each class containing 1,000 training images. After about 2k ...
1
vote
0answers
32 views

Reporting of Neural Network Accuracy for Academic Publications

I'm an academic researcher, working with Convolutional Neural Networks, particularly for image classification. In academic publications, a typical metric for evaluating the performance of a ...
0
votes
1answer
28 views

Need for removing correlated and near-zero variance features despite feature selection?

I'm doing classification with two classes. Before I apply a classifier, I'm doing some preprocessing steps like removing near-zero variance features or highly correlated features (for those ...
0
votes
0answers
50 views

Standardization with mean/std or median/IQR?

I have a dataset with 10000 data points and 20 features. The features are not normally distributed (most of them have a generalized extreme value or burr distribution and all values are greater or ...
0
votes
0answers
22 views

The feature space from Gaussian kernel is infinite-dimensional, are there countably or uncountably many basis?

My attempt: Let $x,y\in\mathbb{R}^d$. We already know the Fourier transform of a Gaussian function is a Gaussian function.If substituting $x-y$ for the variable after Fourier transform, we have $$ ...
1
vote
0answers
13 views

Training set for Bernoulli Process: retain number of “1” examples in proportion to process?

Given a Bernoulli Process, should my training set have a number of "1" examples in proportion to the process? For example, a Bernoulli Process is "1" 10% of the time and "0" otherwise. In a training ...
0
votes
0answers
20 views

Q-Learning: Should I give the sink states a reward, or a Q-value?

I'm implementing Q-learning via neural network to learn the game of Othello/Reversi. Currently, a win gives a reward of 1, a lose gives -1. However, I've run into a dilemma. I don't know whether I ...
0
votes
1answer
30 views

Accuracy on the test set do not change. Why?

I train a SVM classifier using 36 features. If I use all the features, the train accuracy is about 0.96, the test accuracy is about 0.77. Then I change the number of features. The train accuracy drops ...
1
vote
1answer
20 views

Drop in results upon addition of new features in random forest model

I am training a classification random forest for object detection in images. I have several features (like HoG, edge features etc) which work good enough separately. But when I train using all ...
0
votes
0answers
11 views

Influence of correlated features on classifier performance

Let's consider following example. The feature vector has N dimensions. We know that the i feature is linearly correlated to feature j. What we should do in that case. Can we neglect the j-th ...
1
vote
1answer
39 views

Handling missing continuous attribute values in ID3

I'm implementing the ID3 algorithm. I have an attribute which happens to be continuous like 12.21, 3.01, etc. AND have missing values which are marked as "NA". How I'm discretizing the data: I'm ...
0
votes
0answers
30 views

Keras conv-net does not learn anything over epochs

I'm having trouble using Keras to create a convolutional neural network for multi-class classification. The main problem that I'm having is that by altering any one of my: learning rate, momentum, ...
0
votes
0answers
26 views

Using same data twice in a machine-learning model

I am working on a machine learning problem with 37 features to learn from. So the method I plan on using is as follows: 1) I do a sentiment prediction on 17 of these features to output {negative, ...
1
vote
0answers
14 views

Recomender System algorithm for Pinterest-like system

I want to build recommender system for following setup: I’ve got users U Each user U(i) has set of features Fu. Your most common ones – gender, age, country, interests etc. I’ve got collection of ...
7
votes
3answers
386 views

What is the difference between learning and inference?

Machine learning research papers often treat learning and inference as two separate tasks, but it is not quite clear to me what the distinction is. In this book for example they use Bayesian ...
0
votes
1answer
33 views

Evaluate a supermarket recommender system

I have finally come up with a recommender system for the supermarket which now suggests products to users based on implicit collaborative filtering. But I am stuck at a point where I do not know how ...
0
votes
0answers
8 views

Compare the ratio of two variables over a set of sample data?

I have number of applicants and number of candidates recruited for 10000 job openings. I want to interpret the significant difference between two variables. One way is to compare their means. Is there ...
2
votes
0answers
49 views

Recurrent neural network for real-valued prediction with exogenous variables

I have a problem that seems relatively straightforward yet I am stuck on how to proceed. I have several time series of variables $P(t), Q(t), E(t)$ and I want to train an RNN to predict $Q(t)$ given ...
1
vote
0answers
7 views

The estimator of variance of linear regression targets

In Section 3.2 of the book The elements of Statistical Learning (2ed), I read the text: Typically one estimates the variance $\sigma^2$ by $$ \hat{\sigma}^2 = ...
0
votes
0answers
18 views

Supervised semantic analysis

Dimensional reduction and semantic vectorization techniques like LSA, pLSA, LDA and Random Indexing do not take advantage of semantic labeled data like Explicit Semantic Analysis (ESA). I am looking ...
1
vote
1answer
24 views

Classifying unlabeled data, but with cost function

I need to classify objects with ~50 features into 3-4 different classes, there are no labeled examples. Moreover there is no absolutely correct class for any object. However I do have cost value for ...
0
votes
0answers
17 views

R model developing & validating - Open to Discussion [duplicate]

Throughout my R journey I have noticed the way we can use given data to develop and validate a model. Assume that you have given data for a problem train.csv test.csv Method A Combine ...
0
votes
0answers
14 views

Extracting influence counts from Model variables or data

To idetifying the important activity performed from users who have been converted in last N days. So, I have tried GLM, Rpart and Random forest models which can give me the impoprtant activities (in ...
0
votes
0answers
19 views

Do I need a bias for a linear perceptron ? If so how do I update it?

I manage to code a very simple linear classifier, a perceptron, using 2D data. Everything look understandable, it is just that the bias term does not make a lot of sense to me, so I did not use one. ...
0
votes
2answers
51 views

Do we have to scale new unseen feature data for prediction

In machine learning most algorithms require some kind of scaling to decrease error. This is my code: ...
0
votes
0answers
18 views

Feature Selection using chi Square (Accord.Net)

I have data that looks like This is mock data. Let us assume that Sandal,Sneaker,Leather , Boot etc are somehow predictive of 2 classes (male and female) . What is the easiest way to get the chi ...
0
votes
0answers
21 views

How to iteratively learning two matrices in least squares regression?

I have a problem which has the instances (vectors) belonging to two different classes c1 and c2 (same dimensions). I want to learn two matrices M1 and M2, such that sum(M1*x_{c1} - M2*x_c{2})^2 is ...
0
votes
0answers
16 views

Create features from a document

I have been given an assignment related to NLP and I am a newbie in this field. Train a named entity recognition system that treats the documents as strings of mentions, x . A labelling of the ...
1
vote
0answers
22 views

How likely is a hyperparameter search on a subset dataset going to be accurate on the full dataset?

I'm running a hyper parameter search over many options in a neural network, holding all but one fixed in turn until I come to rest on a reasonably optimal set of hyperparameters. I wonder if anyone ...
0
votes
0answers
12 views

Should redundant explanatory variables be discarded?

Suppose that we want to fit a model to predict a given response variable $Y$. Suppose that some explanatory variables are redundant. Be an explanatory variable redundant if it gives similar ...
0
votes
0answers
10 views

pairwise chi square test for independence on large df

I have a data frame (mydata) of discrete data (binned 1 through 10) where each column represents a variable I'd like to use in a Naive Bayes algorithm and each row represents a city throughout Europe. ...
0
votes
1answer
30 views

Possible machine learning methodology to implement both continuous features and features with descrete values in the same model

I would like to implement a machine learning procedure, in order to predict a categorical binary outcome. However, my main concern, is the different "nature" of my features: while a proportion of my ...
3
votes
1answer
46 views

What supplemental resource do you recommend in order to fully comprehend The Elements of Statistical Learning

I am learning the book "Elements of Statistical Learning," but it is very hard because it requires very heavy knowledge about statistics, which I have some, but apparently not enough to understand the ...
1
vote
0answers
24 views

Alternating Least Squares Test Error better than Train

I have been running some trials for recommendations using Collaborative Filtering, specifically Alternating Least Squares (ALS). I am using two versions of ALS, one with fixed lambda regularisation ...
0
votes
0answers
6 views

Measures of multiple-target regression accuracy beyond explained variance and coefficient of determination

I am training a multiple-regression model from a data set of simulations that I generated. I have held-out validation sets where I test how well the model performs. I have calculated the normal scores ...
0
votes
0answers
18 views

Are there simple memory-efficient ways to do multi-instance learning?

At the moment, I'm simply using mean of the features in all the instances in a bag to represent a given bag. I've also tried using min/max, gmean and hmean, but didn't get any better results. Are ...
1
vote
0answers
19 views

Machine Learning - Changes Over Time and Relationship To Inputs

Is there a machine learning algorithm that looks at the change of inputs in relation to each other over time? I apologize for being a bit out of my depth here. I'm hoping for a solution that gives ...
0
votes
1answer
19 views

igraph shortest path export as vector [closed]

This is probably a very simple question but I cannot seem to solve it: I'm using the igraph package and want to export the vpath part of the get.shortest.paths output as a vector so that I may work ...
9
votes
2answers
208 views

Inverse covariance matrix vs covariance matrix in PCA

In PCA, does it make a difference if we pick principal components of the inverse covariance matrix OR if we drop eigenvectors of the covariance matrix corresponding to large eigenvalues? This is ...
1
vote
0answers
39 views

Confidence interval for Bagging method in R

Here is the code which implements bagging that I copied from net (http://www.r-bloggers.com/improve-predictive-performance-in-r-with-bagging/) : ...
3
votes
4answers
609 views

Machine learning methods which takes time-to-event into account?

My vague understanding is that machine learning methods are based on classification labels. How about a survival type of problem? That is to say, not only "have event" or "have no event", but also ...
2
votes
0answers
26 views

how to model longitudinal big data?

Traditionally we use mixed model to model longitudinal data, i.e. data like: ...
0
votes
0answers
46 views

How to calculate accuracy in cross-validation?

I have a classification problem consisting of two classes. I have around 10000 data pionts and 20 features. I'm doing nested 10-fold cross-validation. I am unsure about calculating the accuracy. I ...
2
votes
0answers
44 views

Optimal feature selection

I am working on classification issue. My training set contains of 10D features vectors. As a training model I am going to use Fisher or Neural Network. Here is a plot of the correlation matrix for a ...
2
votes
0answers
28 views

Is it possible to use LASSO regression with multi-levlel data?

I have real-time monitoring data where participants report on a variety of variables four times per day for a month. Is it possible to use LASSO regression (e.g,. glmnet r package) with this data? I'm ...
0
votes
0answers
8 views

trainset and testset with different distributions

I am looking for the research/papers that shows that more balanced train leads to a better macro performance on highly unbalanced test. For example, there is multi-label problem, where 80% of the ...
5
votes
2answers
75 views

Performance benchmarks for MCMC

Have there been large scale studies of MCMC methods that compare the performance of several different algorithms on a suite of test densities? I am thinking of something equivalent to Rios and ...
3
votes
1answer
29 views

Combining transition operators in MCMC

Many MCMC papers usually present a new single transition operator (or a family thereof) such as different proposals for Metropolis-Hastings, new forms of slice sampling, etc. I am interested in ...