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

learn more… | top users | synonyms (1)

0
votes
0answers
23 views

Pre-processing (center + scale, box-cox transformation) inside cross-validation?

I have extracted features and I have now a matrix where the rows are the data points and the columns are the features. Of course, I have to center and scale (zero mean and unit variance) each feature ...
0
votes
1answer
37 views

Proper dataset format for K-Means and DBSCAN clusterers

I'm trying to classify web traffic using clustering algorithms with my own C program, capturing packets with libpcap. In this article K-Means, DBSCAN and AutoClass ...
1
vote
3answers
35 views

Backpropagation: Is there a general weight update rule for both output and hidden layers?

I'm looking for a general weight update rule for both hidden and output layers, no matter the number of layers, the connections or the transfer function. Does anything like this exist? I'm quite new ...
1
vote
0answers
18 views

Should categorical data be normalised in linear regression?

I have data similar to the following: [ [0, 4, 15] [0, 3, 7] [1, 5, 9] [2, 4, 15] ] I used One Hot Encoder to preprocess this data so it is suitable ...
0
votes
0answers
15 views

Which ML technique to use for prediction of right or wrong for a flash card [closed]

Right now I am developing a flash card app that has a machine learning component to it. It should show cards that the user gets wrong more often and cards that users get right less often. I was ...
0
votes
0answers
8 views

what is this technique of training and then having it infer category of object? [closed]

Basically I'm looking for a combination of vision and language as signals to "train" some process where with enough iteration, it is able to "infer" the category of an object. For example, I want to ...
0
votes
0answers
9 views

Create synthetic High frequency data similar to historical real data

The historical data has mean-reverting behavior. Models like OU cannot fully describe hf data if don't incorporate electric jump. Is there way we can create data similar to existing? Maybe ML process ...
-1
votes
0answers
17 views

Automatic mail classification

I'm building a mail classifier in Python 3. I've successfully built classifier to classify spam/ham using SVM (LinearSVC to be precise) using scikit-learn. But the next challenge is to auto bucket the ...
2
votes
1answer
71 views

detect incorrect term in group of terms

I obtain a 'group' of numbers every day. Each number is associated with a 'term'. eg 35 is Big Data. 42 is Hadoop, 82 is Zebra, 89 is Python, 3 is Machine Learning, and 6 is Waterfall, etc. I want a ...
0
votes
1answer
26 views

Machine Learning Procedure for Fractional/Proportional Data?

I am looking for some suggestions of machine learning procedures that work to predict fraction outcomes where the outcome variables $\in [0,1]$. Can you provide me with any suggestions? I thought ...
1
vote
0answers
25 views

Advice for Modeling a Rare Response (y)

My Question 1: Is it possible to create a "good" predictive model with only 40/100,000 rows of data containing a "y" (response) value greater than or equal to 1? My Question 2: Is it a better idea ...
0
votes
1answer
35 views

Regression: find the best degree of polynomial with the best regularization parameter

When trying to predict data using linear regression or classify with logistic regression, with a polynomial, I know how to find the best degree of a polynomial to fits given data when the ...
0
votes
0answers
13 views

Aggregation of Cross-Validated Results

I am using satellite weather features to predict agricultural productivity. I have several models that predict at the daily level. However, I would also like to predict average yield for each week ...
0
votes
0answers
23 views

List of R packages for binary classification that support custom misclassification costs [closed]

I am looking for a list of R packages for binary classification problems that let users set custom misclassification costs for an error on each class. In particular, the packages have to be current, ...
0
votes
1answer
23 views

Comparing kmeans cluster

I have 150 images, 15 each of 10 different people. So basically I know which image should belong together, if clustered. These images are of 73 dimensions (feature-vector) and I clustered them into ...
0
votes
1answer
20 views

How to train and fine-tune fully unsupervised deep neural networks?

In scenario 1, I had a multi-layer sparse autoencoder that tries to reproduce my input, so all my layers are trained together with random-initiated weights. Without a supervised layer, on my data this ...
2
votes
0answers
21 views

How do I calculate the derivative of kurtosis and entropy?

I'm using kurtosis and entropy as penalty terms in my neural network's cost function. Need to back-propagate the error. For that I need a quick way to estimate the derivative (the gradient ...
0
votes
1answer
10 views

Is there a framework for reinforcement learning with states and actions in the same domain?

In reinforcement learning, there are states, actions, initial states, terminal states, a progress function and a reward function. Is there a theoretical framework or setting where states and actions ...
0
votes
0answers
15 views

Problem with SVM methods on mathematica [closed]

I am running SVM to do human activities classification. Vector features is an histogram which represents a probability distribution (Total[feature vector]=1). However, values in features vector are ...
1
vote
0answers
39 views

Is learning in statistics = hyper-parameter/model-learning in ML and inference in statistics = learning in ML? [closed]

It seems that when statisticians refer to learning, ML researchers refer to hyper-parameters and model-learning, and when statisticians refer to inference, an ML researcher refers to learning the ...
0
votes
0answers
40 views

Multi label image classification using convolutional neural network in Python

I am working on multi label image classification problem. The dataset is given on this link. I am using Convolutional Neural Network (CNN) with fully connected neural network (NN) at the end. I am ...
0
votes
0answers
13 views

Assessing significance / confidence of a crossvalidated performance measure

I have a prediction model $P$ and I use some performance measure $I$ to measure $P$'s accuracy. The distribution of $I$ is unknown (it's a custom metric, which is somehow similar to the precision ...
0
votes
0answers
19 views

Difference between feature, feature set and feature vector

I learnt that a feature is an individual measurable property of a phenomenon being observed. Say for example, I am representing a human being. Then various features could be the age, weight, ...
0
votes
0answers
20 views

Algorithm question [closed]

Kindly help me with this question: • A short design document detailing what you and why, which algorithms were used and what methods. • Working code solving the problem, and sample output for the ...
0
votes
0answers
28 views

Online Machine Learning of sequential events with varying delay

Lets say we have A to Z features which repeat sequentially. So you have A(1), B(1), ... Z(1) at time 1 followed by A(2), B(2),....Z(2) at time 2 and so on till A(n), B(n), ... Z(n) at time n. Each of ...
0
votes
2answers
27 views

Using advance optimisation techniques for collaborative filtering systems, is it possible?

Is it possible to use advance optimisation(L-BFGS, Conjugate gradient) for an collaborative filtering systems vs just using gradient decent? I ask this because of the need to calculate both X and ...
0
votes
0answers
22 views

How to compare the feature importances produced by two different classifiers?

In one study, I am using two different classifiers. I want to compare the feature importances produced by two classifiers. Is there a statistical technique to measure the similarity between the two ...
0
votes
0answers
16 views

Framework/tool for scalable automatic model training in batch and online?

For a retail problem, we need to build model for every individual customer on their accumulated historical data on a periodical basis. This is to predict certain classes individually. For a ...
1
vote
0answers
16 views

How should I transform a featureset (15000 variables) that is mostly presence/absence, but present values are lognormal distributed?

I am trying to learn machine learning and have a nice featureset with a binary classification. The dataset is 15000 variables and 2500 data rows. For every data row, almost all variables are 0, and ...
0
votes
1answer
36 views

How to interpret my learning curve

I created the following learning curve in order to diagnose my Random Forest model. As I can see the curve indicates high variance and 'underfitting' (not overfitting), because cross-validation ...
1
vote
0answers
20 views

Using an RNN/LSTM to generate sequences with a unique output

I'm trying to train a LSTM recurrent neural network where my data consists of a sequence of animal migration data ...
0
votes
2answers
51 views

Dimension reduction using PCA in Matlab

I have a $152 \times 27578$ matrix, $152$ samples and $27578$ features, and I used the PCA function for the dimension reduction in Matlab. ...
0
votes
0answers
20 views

What about transforming autoencoders allows them to learn pose parameters?

After reading several papers on transforming autoencoders and looking through an implementation on github I'm curious what it is that enables them (and not standard autoencoders) to learn pose ...
2
votes
1answer
23 views

When implementing dropout in neural networks with SGD, how does one calculate the gradient?

Specifically, I know that in SGD one sums all the gradients for weights/biases for each minibatch and divides by the mini batch size, would one do the same thing for dropout networks? Or would they ...
1
vote
1answer
63 views

Predict time of next purchase

I'm trying to build a model in R that will let me predict when a costumer will purchase a product again. For example, the training data list customers who purchased bikes. I want to predict when ...
-1
votes
1answer
22 views

Big data set for Document Classification [closed]

I'm looking for big data set which is suitable to be used for document classification task. The data set which I'm looking for should composed of the frequency of the words which exist in each ...
0
votes
1answer
69 views

Is there “rule of 30” in machine learning

In the Deep Learning course on Udacity, the instructor mentioned "rule of 30", means that your new method / algorithm ... can be considered as "significant improvement" if it can improve the results ...
0
votes
0answers
31 views

Boruta feature selection with R

I have a concern :/, I used Boruta algorithm in R on a data set of ~600 attributes and with a sample of 50K (the original size is 300K). Using the following parametres: pValue = 0.05 getImp = ...
0
votes
1answer
39 views

LSTM tutorial question about dropout

I have a question related to the dropout function in the LSTM tutorial: http://deeplearning.net/tutorial/code/lstm.py ...
1
vote
1answer
30 views

What prevents duplication of neuron parameters in fully connected layer

LeNet has several fully connected layers, I'm wondering what prevents neurons duplicate other's weights and outputs. Unfortunately the only technique, I came up with, is random weight initialization ...
1
vote
0answers
23 views

grouping attributes in RF and GBM

i have a dataset with 1000 samples and ~11k features (SNP markers). i have identified 100 additional binary features describing the markers themselves so i have a ...
0
votes
0answers
98 views

Email open-rate optimization

I am trying to maximize open rates of emails by selecting between two subject headlines {h1, h2} for a marketing campaing. The hypothesis is that different customers react to different headlines. ...
0
votes
0answers
19 views

R C5.0 tree model to list conversion

I am using the C5.0 decision tree in R from the C50 package. The training function C5.0 returns a list which also contains a "tree" element which is basically a text representation of the tree. I am ...
0
votes
0answers
35 views

omit variable to find feature importance in classification

I have not seen the following idea about measuring feature importance in classification discussed. Fit model with all variables; get full model classification accuracy Successively fit the model ...
0
votes
0answers
31 views

Decision Theory

Patient X is worried that he may have disease Y. He goes to a doctor who performs some test and based on the test determines that the probability that X has disease Y is 0.3. The insurance company has ...
3
votes
1answer
50 views

Clarification about no free lunch theorem

So I initially thought that the NFL theorem meant that an algorithm that is good at learning in one problem domain is necessarily bad at learning in a different problem domain. But, after reading a ...
3
votes
0answers
31 views

Cut off point with three categories

I have a categorical variable called severity with three categories (low, medium, high) and another numerical variable call ZX that can take values from 0 to 10. I want to find the cut off points of ...
1
vote
0answers
19 views

Is the Laplace/Lidstone smoothing parameter (talking about Multinomial/Bernoulli Naive Bayes) related to the particular structure of the dataset?

I'm working with Multinomial and Bernoulli Naive Bayes implementation of scikit-learn (python) for text classification. I'm using the 20_newsgroups dataset. From the scikit documentation we have: ...
1
vote
1answer
40 views

Neural networks with complex weights

I am currently wishing to give a neural network starting weights with complex values (because of the nature of the specific task I am working with). I was trying to use the standard neural net ...
0
votes
0answers
28 views

Is feature complementarity different from feature interaction?

I am writing a conference paper in which I have a sentence like "...complementary/interactive features...". This sentence ...