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

learn more… | top users | synonyms (1)

0
votes
0answers
17 views

What exactly is a background class in a classification problem?

I have heard people using the term background class in the following to scenarios: For a class which has a very high number of instances compared to other classes in a classification problem ...
0
votes
0answers
10 views

Why is my high degree polynomial regression model suddenly unfit for the data?

I'm building a ridge regression model in scikit-learn and trying to find the optimal degree polynomial to use. The data I'm working with is a fairly predictable time series of hourly traffic volumes, ...
1
vote
3answers
482 views

Implementing WARP Loss (Gradient Computation)

I am trying to implement the WARP Loss in Torch, as defined in the WSABIE paper: http://www.thespermwhale.com/jaseweston/papers/wsabie-ijcai.pdf The Algorithm is as follows: The Algorithm ...
-1
votes
0answers
28 views

By use NaiveBayes, get result in probabilities (0.5, 0.7) not only in 0 and 1

I have tried to find an answer, but could not. If I use NaiveBayes on train data after find optimal settings apply it on test data. but the result which U receive is 1 or 0. By using DISCRETIZER I ...
7
votes
1answer
81 views

What are the impacts of choosing different loss functions in classification to approximate 0-1 loss

We know that some objective functions are easier to optimize and some are hard. And there are many loss functions that we want to use but hard to use, for example 0-1 loss. So we find some proxy loss ...
0
votes
0answers
19 views

Feature scaling of non-Gaussian data before SVM

I have data for a binary classification problem and was wondering generally what to do if the different dimensions/features of your input training data display vastly different distributions in ...
1
vote
3answers
648 views

Differences Between Logistic Regression in Statistics and in Machine Learning

I just found out that machine learning also has logistic regression as one of its methods. Can someone please tell me the differences between logistic regression in statistics and machine learning? I'...
2
votes
1answer
20 views

multivariate time series clustering

I am collecting a group of multivariate time sequences. For example, there are 2000 time series. Each time series is of 12 dimensions. Are there any systematic models/algorithms that can cluster ...
0
votes
0answers
7 views

How to design RNN\VAE for sparse sequential data?

I am currently struggling with designing some Deep Learning algorithms that takes as an input sparse sequential data, since the system I try to model signals sparsely (once in a while). I try to ...
4
votes
3answers
1k views

“objective function, cost function, loss function” Are they the same thing?

In machine learning, people talk about objective function, cost function, loss function. Are they just different names of the same thing? When to use them? If they are not always refer to the same ...
3
votes
1answer
364 views

Comparison of two classifiers based on precision/recall/F1 only?

For two classifiers h1 and h2 I have the precision, recall and F1 score as a percentage (along with the original labeled data set that they were tested on). If I had access to which samples each ...
0
votes
1answer
34 views

Why is “the true data generating process” similar to simulating the entire universe?

In the deep learning book by Bengio, Goofellow and Courville (http://www.deeplearningbook.org/) there is paragraph in the regularization chapter. "Deep learning algorithms are typically applied to ...
0
votes
0answers
18 views

When doing regression with a singled layered Neural Network, what activation function is the best one to choose?

I was training a singled layered (shallow) neural network as in: $$ f(x) = \sum^K_{k} c_k\theta(W_k x+b_k)$$ for regression (using squared error loss) or function approximation. I was wondering, is ...
0
votes
0answers
20 views

Time Series Prediction using Machine Learning

I am trying to predict the request arrive time of some objects in network traffic. I have few features of the object like their type, size, previous arrival time, etc. So I was think that I should use ...
1
vote
0answers
39 views

Time Series Data Mining and Correlation guidance [on hold]

Introduction Editing this post to clarify what I am asking. This is a bit of a vague question to begin with. I am not asking if this particular statistics method with these variables will cause this ...
-1
votes
0answers
17 views

What models are well suited to varying-length feature vectors?

I'm trying to classify a bit of a strange-looking data set where the input vectors have a variable length. The actual data is pretty boring, but imagine a related problem where you're trying to ...
0
votes
0answers
16 views

Sparsity as missing data + MLE

I just had this "funny" idea: what about a classifier that not only tries to learn weights for predicting y but actually works with "deleted" data (as in sparse ...
3
votes
2answers
382 views

Bayesian Linear Regression

I have the following question concerning Bayesian linear regression on my machine learning assignment: Consider $f = w^Tx$, where $p(w) ∼ N(w | 0, Σ)$. Show that $p(f | x)$ is Gaussian. I ...
2
votes
1answer
35 views

Determine the feature weights with a regression?

I have a set of houses with their features (location, size, number of rooms, etc … and the y is the price). In the future I will have a new house without the price. My goal is to find the 20 closest ...
1
vote
1answer
28 views

How does data augmentation reduce overfitting?

I'm trying to understant the benefit apported by the step of data augmentation in a classification algorithm. I have a vector of hexadecimal strings and a column vector containing the label ...
12
votes
1answer
6k views

Supervised learning, unsupervised learning and reinforcement learning: Workflow basics

Supervised learning 1) A human builds a classifier based on input and output data 2) That classifier is trained with a training set of data 3) That classifier is tested with a test set of data 4) ...
1
vote
1answer
92 views
+50

Can I hack weighted loss function by creating multiple copies of data

Suppose we want to build a binary classifier with weighted loss, i.e., it penalize different types of errors (false positive and false negative) differently. At the same time, the software we are ...
0
votes
1answer
15 views

How to get from input depth to output depth in convnets?

See this example: convnet quiz Udacity. How to get from input depth = 3 to output depth = 8? My assumption: In this example we have 8 filter (kernels) and each of them slides over the 3 inputs. So in ...
0
votes
2answers
158 views

highly sporadic validation error during training with multilayer perceptron

I'm encountering an issue where a classifier I'm developing reports validation errors during training that span a wide range of values without consistently decreasing over time. Unfortunately, I'm new ...
0
votes
0answers
13 views

RMSD vs. Log-loss

I have four vectors of numbers, one is the ground truth (binary, rach number is either 1 or 0), the other three are the corresponding prediction (each number is a probability from 0 to 1) generated by ...
3
votes
1answer
1k views

Clustering text with python

I have asked on StackOverflow, but they suggested me to move here for better answers. I copy paste the question. I have decided to play a little with similarities and clustering text. I have already ...
0
votes
1answer
18 views

Classification: training sets different sizes

I'm building a classifier for text analysis sentiment. I have a large training set for positive, neutral and negative mentions. Should the training data sets be similar in size? Currently my ...
1
vote
1answer
31 views

Should I optimize neural networks that are part of ensemble of neural networks?

I'm creating ensemble of neural networks for a simple binary classification task. Every neural network is generated and trained a bit differently (number of hidden layers, number of neurons per layer, ...
2
votes
2answers
752 views

Cross validation procedure - is this right?

Just want to check that I am performing my cross validation procedures right. I'm using a non-linear svm. I do a five fold cross validation (5 splits of test/train on my original training data) and ...
1
vote
0answers
5 views

Time interval prediction

I want to predict a specific time interval(ex. patient processing time in clnic) with some boolean value like whether this patient has cough or whether he/she has certain disease. I have tried using ...
0
votes
1answer
129 views

Linear regression to minimize the Cost Function

I am new started machine learning program. I find it difficult to understand gradient descent algorithm. I am going through machine learning from coursera by Andrew Ng.All of his lecturer in second ...
0
votes
0answers
7 views

If number of examples is smaller than number of features, but greater than size of my filter, is my convolutional network still okay?

I'm training a 5-layer convolutional neural network on about 300 images with size $30 \times 20$. We can say that the number of features before the first layer is $600$. So it seems like I don't have ...
28
votes
3answers
3k views
+50

Understanding Naive Bayes

To demonstrate the concept of Naïve Bayes Classification, consider the example displayed in the illustration above. As indicated, the objects can be classified as either GREEN or RED. My task is to ...
-1
votes
0answers
17 views

Data science : where to find Project Examples using Python with SQL? [on hold]

I’ve learnt Python and SQL, but was suggested to “work in Python to analyze data sets stored in MySQL”. I am following CS109 which uses Python, but it does not really use SQL https://www.quora.com/...
0
votes
1answer
12 views

Prediction of next event occurrence

I am working on a request prediction problem in which I have to predict which object will be requested in the near future and how many times. This is like a basic internet traffic request pattern on a ...
9
votes
3answers
3k views

Why do we divide by the standard deviation and not some other standardizing factor before doing PCA?

I was reading the following justification (from cs229 course notes) on why we divide the raw data by its standard deviate: even though I understand what the explanation is saying, it is not clear ...
1
vote
1answer
35 views

Ensembles - few questions on approach with multiple models

I'm looking for some general-high level understanding of how I should apply ensemble techniques. I understand that some models can be already thought of as ensembles - such as random forests or some ...
-1
votes
0answers
69 views

Prisma App Neural Networks [on hold]

Can anyone point me towards relevant material that would allow me to understand how Prisma works? http://prisma-ai.com/index.html
0
votes
1answer
27 views

Bootstrapping test set?

Let's say I have a classification problem with a small and fixed test set. If I train a classifier and report the accuracy on this test set, I know that this estimate has a high variance. Does it make ...
1
vote
1answer
32 views

Regression + Time series

I have time series data about sales/day, but i also want to include other data (static/dynamic) to forecast the time series. Is it possible to combine ARIMA model and regression models to achieve the ...
0
votes
0answers
25 views

Ordering by importance when plotting partial functions from forestFloor

I would like to start by saying that I am relatively new to the R packages randomForest and forestFloor and I have never posted a question on here - so bare with me. I have been working on an ...
3
votes
0answers
32 views

Family of flexible parametric mappings $f_\theta:(0,1) \rightarrow \mathbb{R}$?

For the purpose of reparameterizing a model (mostly with the goal of improving MCMC efficiency), I am looking for a family of flexible parametric mappings $f_\theta:(0,1) \rightarrow \mathbb{R}$ such ...
0
votes
0answers
18 views

Test Support Vector Machine on its own Training Data: issues and observations

I recently concerned myself with SVMs and Sparse Bayesian Machines (RVMs) and was performing feature scaling of my highly skewed data prior to the training of my algorithm. When I was subsequently ...
1
vote
0answers
22 views

Training instances importance in Random Forest?

Is it possible to determine the importance of the training examples in Random Forests, analogously to what's done with predictors? Basically the idea would be to find important samples in the data, ...
0
votes
0answers
40 views

how to preprocess/feature stacle multimodal input data?

I am wondering how to normalize data for the use of SVMs etc. that has a clear non Gaussian, i.e. non unimodal distribution. I wrongfully scaled the data by subtracting the mean and dividing the std ...
1
vote
1answer
282 views

Collaborative Filtering for Implicit Feedback Datasets

I'm building a recommendation engine using ALS, as described in the title document. I'm confused about a few points: How should one interpret both X&Y, where X and Y are "factor vectors in the ...
1
vote
0answers
21 views

Interaction between two random variables [on hold]

I have a machine learning system called Sys. Sys uses a variables set (Var) applied to a ...
2
votes
1answer
65 views

Which classification techniques perform efficiently under homomorphic encryption

I am reading a paper (pdf) on homomorphic encryption and its use in machine learning. This paper explores classification methods like Fisher Linear Discriminant Classifier (FLD) and the Linear Means ...
0
votes
1answer
220 views

featurizing images of different sizes

I'm training a non linear svm to do classification on images. I'm featurizing the image by creating 3 features for each pixel, its rgb value. My question is: How should i normalize images of different ...
2
votes
0answers
15 views

How can I apply discriminant analysis (multivariate) if my data is not normally distributed?

As LDA and QDA assume normality and homocedasticity, and my data is not that well behaved... is there any other kind of robust technique I could use? Could applying box cox transformations and then ...