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

learn more… | top users | synonyms (1)

0
votes
2answers
29 views

choosing a model after feature selection process

so ive been selecting features for a regression problem and have obtained a list of the best performing feature sets. (note my list is actually several thousand lines long) 188.493 186.989 [379.45, 0....
0
votes
0answers
13 views

Marginalized Denoising Autoencoder for Regression/Prediction

I have been looking online at papers etc. about marginalized denoising autoencoders (mda) and everything I've found so far uses mda for pre-training layers for a classifier such as a support vector ...
0
votes
1answer
24 views
0
votes
0answers
67 views

In linear regression, how do I extrapolate parameters obtained using preprocessed data?

Where $h_{\theta} = \theta_{0} + \theta_{1}x$, I am trying to minimize $J(\theta) = \frac{1}{2m}\sum_{i = 1}^{m}(h_{\theta}(x^{(i)}) - y^{(i)})^{2}$ I first transform every sample in the feature (...
0
votes
0answers
18 views

Advantage of character based language models over word based

Is there an intuition why character based models language bases models are preferred over word based. For example Karpathy builds his language model by predicting the next character in Karpathy Blog. ...
1
vote
2answers
87 views

Speed of prediction: neural network vs. random forest?

I'm currently trying to improve on a classifier. The current method used is a neural network, and the method I've found to be better is a random forest (or even just a single tree). With 40 trees, the ...
0
votes
0answers
11 views

What happens to the RKHS of a Gaussian kernel if $\sigma$ is increased / decreased?

I'm especially interested in the following case: let's say we have a Gaussian kernel $K$ with bandwidth $\sigma$ and RKHS $\mathcal{H}$ and a set $H = \{ \forall h \in \mathcal{H} : ||h||_K \leq \...
0
votes
0answers
16 views

Clustering for mixed variable type [duplicate]

I have the following data set with mixed variable types ...
0
votes
0answers
23 views

what is better for demand forecasting, NN or ARIMA?

I want to predict future demand based on data I have. I hava already used ARIMA model but it is not giving very high accuracy. So should I also try NN or ARIMA is better?
0
votes
0answers
6 views

How to segment hours of audio for speech recognition?

I have 36 hours of speech data along with transcription. I'm planning to have 7 second audio segments, because I don't know any better. Suggestions are welcome. These segments will be passed through ...
0
votes
0answers
9 views

Labelled data labels are changing in sklearn Label Propagation

0 down vote favorite I have some words which have labels of '1' or '0' and '-1' for the words that I don't have labels. And when I run Label Propagation algorithm of sklearn,it is even changing my ...
0
votes
0answers
40 views

Using machine learning to detect errors

I have a scenario I believe could benefit from some of the statistical models used in ML but I need a little guidance. Or someone to tell me I'm way off base with my idea. The scenario: I have ...
0
votes
0answers
12 views

Improve Chan-Vese algorithm by machine learning

Chan-Vese Algorithm is an unsupervised image segmentation method. It finds the boundary curve $C$ by minimizing the object function $$ F(c_1, c_2, C) = \mu \cdot \text{Length}(C) + \nu \cdot \text{...
1
vote
0answers
10 views

constant terms in stochastic gradient descent: when to apply them and how much of the constant gradient component?

in a derivation for the gradient of a collaborative filtering system (similar to Probabilistic Matrix Factorization), I got to the following expression for the gradient of a latent vector $\mathbf{u}...
0
votes
0answers
9 views

Relation between number of states and Gaussians in HMM GMM chain and window size of Feature Extraction Algorithm

I am using HMM-GMM tool for Matlab by Kevin Murphy.My Frame work has a feature extraction Algorithm (MFCC,Spectrogram) followed by a HMM-GMM classifier. I am trying to set the the number of states ...
1
vote
1answer
50 views

Distance between rankings - Kendall tau

I want to use Kendall in order to measure the distance between rankings A and B. How do we deal with different ranges? For example, as shown below: let's say we have the following rankings A and B - ...
8
votes
4answers
545 views

Gradient boosting machine accuracy decreases as number of iterations increases

I'm experimenting with the gradient boosting machine algorithm via the caret package in R. Using a small college admissions dataset, I ran the following code: <...
4
votes
1answer
103 views

outlier detection: area under precision recall curve

I would like to compare outlier detection algorithms. I am not sure if area under roc or under precision recall curve is the measure to use. A quick test in matlab gives me strange results. I try to ...
3
votes
1answer
67 views

Machine Learning Algorithms for Predicting Entire Regression Functions

From what I've seen in (supervised) machine learning the general idea is to work with some training set $(\mathbb{x_i},y_i)$ and learn the $y_i$ outputs without over-training. In the case of ...
1
vote
0answers
30 views

SVM subgradient-descent (Pegasos algorithm)

I am trying to implement the Pegasos algorithm for large scale SVM training. I'm following the main paper Pegasos. Everything worked fine but the results are quite disappointing. The code: ...
0
votes
0answers
15 views

Creation of Labels for Label Spreading in skicit-learn

I am using Label Propagation in skicit-learn for finding labels for the unknown ones My Input is 'data_list' containing 400 000 sentences in sanskrit language like : ['tatra yad tad mahABAga SaMkara ...
0
votes
0answers
24 views

Multiclass using binary classifiers in R

I am dealing with a problem where I have about short pieces of text (around 500 characters) and about 70 class labels. Currently each text is only assigned to one class, but we'd like to extend it so ...
7
votes
2answers
322 views

Do all machine learning algorithms separate data linearly?

I am an enthusiast of programming and machine learning. Only a few months back I started learning about machine learning programming. Like many who don't have a quantitative science background I also ...
0
votes
0answers
14 views

SVD - collaborative based filtering - Prediction matrix

On the movielens dataset, I used SVD to find U, s, and V matrices. Then performed the dimensional reduction by elimination of everything corresponding to lower valued eigen values( upto a threshhold). ...
0
votes
1answer
34 views

Is there any ranking neural network especially in R? [closed]

I know there are many neural network for classification and regression problems. I myself use caret package in R for those problem. But now I am looking for a neural network which can be used for ...
0
votes
0answers
16 views

How to asset bias in data used to update a recommender systems?

I want to study the bias in a recommender systems.So,in each iteration,the recommender systems update the model using the coming data(new ratings) from users.and then, the RS recommend a top N items ...
1
vote
0answers
29 views

Counter intuitive behavior from scikit-learn's SGDClassifier

I am working with SGDClassifier from Python library scikit-learn, a function which implements linear classification with a ...
1
vote
0answers
18 views

Factorization Machine Algorithms And Implementations For Implicit Feedback?

Spark ASL supports only (user, item, measure) implicit pairs, libfm supports any number of features but no implicit feedback ranking (only classification/regression). Is there good articles/...
3
votes
1answer
32 views

look up table as a special case of a the linear function approximation (Reinforcement learning)

In reinforcement learning, where the state space is discrete and relatively small, a form of learning algorithm commonly used is the Q learning. This involves a look up table $Q(s,a)$ (Or you think of ...
0
votes
1answer
33 views

Cross entropy-equivalent loss suitable for real-valued labels

I am building a model whose outputs are between 0-1 and the goal is to minimize a cost function over the predicted values and labels. So far everything seems to be easy but my labels are real-valued ...
0
votes
0answers
8 views

Score fusion vs Stacking

I was reading a post that used score fusion to compare two scores from two different classifiers (after normalisation). I read another that suggested feeding the results of these two classifiers into ...
1
vote
0answers
42 views

What is an explanation of the example of why batch normalization has to be done with some care?

I was reading the batch normalization paper[1] and it had one section where is goes through an example, trying to show why normalization has to be done carefully. I honestly, can't understand how the ...
0
votes
0answers
45 views

how to measure similarity of two datasets (matrices) of different length

There are related questions being asked already but my problem is i can't find a good method of measuring similarity between two datasets that are represented by various lengths of matrices. For ...
0
votes
0answers
9 views

Why am I getting the same value for F1 and accuracy?

I trained and SVM classifier and I noticed that I'm getting equal F1 and accuracy values (using a cross-validation), which means that the number of True-Positives and True-Negatives is the same. The ...
0
votes
0answers
7 views

Spatial structure of CNN features

Is there any work to study the space of features learned by convolutional neural networks like if they perhaps lie on a manifold? What about other representations like SIFT?
1
vote
4answers
112 views

Which book should I read to get started with machine learning, Elements of statistical learning or Pattern recognition in machine learning?

I want to learn machine learning. I found tons of material on the internet but couldn't decide which book to get started with.
0
votes
0answers
25 views

Intuition behind error updating for inner layer neurons

In backpropagation the usual way to calculate the amount of error of a layer is $$\delta_0 = y_{expected} - y\\ \delta_i = \sigma'(input)\sum_{j \in outputs(i)}{\delta_jw_{i,j}}$$ where $\sigma$ is ...
0
votes
0answers
20 views

Classification of real values

I have values of attribute between 0 and 1 which i want to predict. The distribution of values is shown in fig. I want to predict this attribute. The problem is there are around 15 classes in this ...
0
votes
0answers
33 views

Identify outlier usage intervals in time-series data

I want to find outliers in power consumption in real-time, at hourly rate, i.e., at the end of the hour, I should say whether power consumption in current hour was outlier/anomalous or not. Approach: ...
0
votes
2answers
61 views

Why Does a Variable with Weak Correlation to Outcome Variable Emerge as Most Important Factor in Random Forest?

I'm puzzled about why a dependent variable with the weakest correlation to the outcome variable emerges as the most important factor when I run my random Forest on the same dataset. It beats out ...
0
votes
0answers
14 views

Statistically significant increase in Matthew's Correlation Coefficient

Suppose I have two classification algorithms A1 and A2, and a test set of size $n$. I evaluate A1 and A2 on this test set, and get corresponding Matthew's Correlation Coefficient scores $\textit{MCC}...
1
vote
2answers
325 views

I did PCA of my dataset with two classes and here is the scatterplot; how can I tell if my dataset is learnable?

What should I look for in my PCA? I'm doing supervised learning with (unfortunately only) 2000 examples, evenly split into 1000 yes and 1000 no. Each vector is a 1000 dimensional boolean vector. I ...
0
votes
1answer
42 views

Machine Learning - How to use a classifier to find the most likely model

I have been learning about the use of machine learning algorithms and their application to particle physics. Now, I have some doubts concerning what to do with the results. Let me explain: imagine ...
0
votes
2answers
39 views

How can I find the field which most affects or contributes to decision making in a machine learning algorithm?

Consider the example below. On a larger dataset, it would be fairly obvious that name and gender are not a good indicator of whether a person is an adult or a kid, and that it's age which best decides ...
1
vote
0answers
23 views

What is the best form (Gaussian, Multinomial) of Naive Bayes to use with (one-hot encoded) features?

I've been asked to use the Naive Bayes classifier to classify a couple of samples. My dataset had categorical features so I had to first encode them using a one-hot encoder but then I was at a loss ...
0
votes
2answers
20 views

Ranging [0,1] test set with parameters from training set

I am working on Machine Learning, particularly I have a dataset with 50+ columns and 100,000 rows. I need to get the data normalized with ranging to [0,1] (not with standardization) and I've split the ...
31
votes
6answers
1k views

Variable selection for predictive modeling really needed in 2016?

This question has been asked on CV some yrs ago, it seems worth a repost in light of 1) order of magnitude better computing technology (e.g. parallel computing, HPC etc) and 2) newer techniques, e.g. [...
0
votes
0answers
24 views

kNN bad performance for iris data set

I've implemented kNN algorithm in Python and now I'am testing it on iris data set. I have two questions. The performance seems to be bad: if I run the program 100 times and then calculate the ...
0
votes
0answers
8 views

How to select weights in the F-Measure to be aligned with that used in cost-sensitive SVM training?

I am dealing with a classification problem in which Recall is more important than Precision, and the training dataset is an imbalanced one. The approach I am taking is to use oversampling to mitigate ...
1
vote
1answer
16 views

Use case for Anomaly Detection using the multivariate Gaussian distribution

We have 5000 vehicles of different classes (trucks, small cars, large cars) with 100 sensors in each car measuring fuel consumption, distance traveled, average speed etc for some time period $t$ that ...