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

learn more… | top users | synonyms (1)

0
votes
0answers
22 views

Can anything be cost function? [duplicate]

I've asked already here but didn't get any answer. If we have some activation function and cost function in neural network, weights will be updated depending on derivation of those functions. With ...
1
vote
3answers
96 views

Which is the best classifier and with what performance measures?

I tried to implement a Classifier comparison like in the scikit-learn for text classification. I used an 81 instances as a training sample and a 46 instances as a test sample. I tried several ...
3
votes
2answers
114 views

RMSE - where this evaluation metric came from?

Does anyone know where this metric came from ? Can someone bring article references or something like this? Im actually wondering if there's any mathematical concept or any way to demonstrate ...
0
votes
1answer
48 views

When to use ANN with tensorflow?

I'm new to machine learning and tensorflow and I'm confused as to why (and when) to use the types of ANN (ie recurrent neural network) with tensorflow? I know RNN is good for sequences of data/time ...
2
votes
0answers
30 views

Is stagewised feature engineering/ selection an invalid approach?

Suppose we want to build a regression or classification model. However, the features (independent variables used) are not all ready at one time. This is very realistic in business, because the data ...
0
votes
1answer
21 views

Is there a heuristic for determining the size of a fully connected layer at the end of a CNN?

For example, in VGG/OxfordNet, the fully connected (dense) layers that precede the final classification layer are of size 4096. Similarly, in an AlexNet ... the number of neurons in the ...
4
votes
1answer
93 views

Is there a supervised learning method for classifying graph walks?

I have geolocational data(coordinates and times with device id), I can bucket this down using say 5m by 5m squares to represent a vertice on a graph. Then following the device id and creating edges ...
1
vote
0answers
34 views

Derivation of derivatives in backprop stage of convolutional neural network

I'm reading the online book "Deep Learning" by Ian Goodfellow et al. . In section 9.5 Variants of the basic convolution function: Directly quoting from the book, "Suppose we want to train a ...
0
votes
2answers
44 views

Should using SMO classification in WEKA take so long with large dataset?

I have a dataset of 205 features and 238000 samples. It is a combined dataset of several subjects' data that I want to use for between-subject classification. I am using WEKA 3.8 with a 64-bit JVM ...
1
vote
2answers
43 views

Interpreting multinomial logistic regression in scikit-learn

I am running a multinomial logistic regression for a classification problem involving 6 classes and four features. Here is the code: ...
0
votes
1answer
45 views

Confused about logistic regression equality

Problem: Prove that: \begin{align} \Delta E(in) &= -\frac{1}{N} \sum_{n=1}^N \frac{y_n x_n}{1 + e^{(y_n w^t x_n)}} \\[10pt] &=\frac{1}{N} * \sum_{n=1}^N - y_n x_n \theta (-y_n w^T ...
2
votes
1answer
77 views

relationship between fully connected layer and convolutional layer

When reading about the transforming the fully connected layer into convolutional layer, posted in http://cs231n.github.io/convolutional-networks/#convert. I just feel confused about how to understand ...
0
votes
0answers
24 views

Single pass object detection

Let we have a set of images $\{\mathcal I_i\}_{i=1}^n$ with labels $\{\mathcal B_i\}_{i=1}^n$, where each $\mathcal B_i$ is a set of regions. The problem is to find a function that given image $\...
0
votes
1answer
32 views

Calculate EER from FAR and FRR?

I'm wondering if we have FAR and FRR scores for each threshold if we can compute an EER programatically? Say we have: ...
0
votes
2answers
23 views

Neural network input values belonging to classes

I need help on configuring a neural network. I would like to pass in accelerometer values (x,y,z) from two different sensors, and have the network compute the corresponding angle. I am providing close ...
1
vote
1answer
26 views

Are there disadvantages using proportional features instead of absolute values?

I was wondering whether there are disadvantages in using proportional features instead of features with absolute values. For example: I have the following data set, which includes, TV duration, ...
0
votes
0answers
44 views

Creating training data for machine learning

I am creating a machine learning model but I don't have any data as such that can be used as training data. All I now that there are certain independent features/columns(categorical and numeric both) ...
1
vote
1answer
48 views

Why are the weights of RNN/LSTM networks shared across time?

I've recently become interested in LSTMs and I was surprised to learn that the weights are shared across time. I know that if you share the weights across time, then your input time sequences can ...
1
vote
1answer
31 views

Bottleneck building block in Residual learning networks

I am wondering about how 1x1 convolution can be used to change the dimensionality of feature maps in a residual learning network. Here the top 1x1 convolution changes the feature map size from 256 ...
2
votes
0answers
37 views

How to improve the sensitivity of minority class on imbalanced datasets

I am working on a classifier which stratifies a population of samples into different classes. The class distribution (ground truth) is imbalanced, and the prevalence of each class is: $$\begin{...
-1
votes
1answer
17 views

can ML learn concepts in an unsupervised manner

I am wondering if ML can currently learn concepts in an unsupervised manner and how would that work. For example, when it looks at a transaction, I would like ML to understand the concept of the ...
0
votes
0answers
18 views

knn text clasiffication model error when term is not found in new documents

Using KNN model for topic clasification. My model uses 200 variables (terms) and 10 target labels, using R (tm package). Accuracy is fairly good. Now, new documents are arriving that need to be ...
0
votes
0answers
51 views

Cost functions like cross-entropy in backpropagation for non sigmoid activation?

I was following this resource. Cross-entropy function was introduced as cost function. When calculating gradient in backpropagation we get delta values which depend on derivation of activation ...
0
votes
1answer
30 views

How to calculate average EER and ROC?

If I test a system on 3 users and I obtain FARs and FRRs for thresholds and an EER for each user, how do I obtain an average EER and plot a ROC? As I see it, I could average the 3 EERs I have ...
0
votes
0answers
27 views

Machine Learning: Non-Linear Regression over dataset with very similar predictors and very different targets

I have a time-series dataset collected by a group of biologists counting the abundance of a particular animal species in an area. I later enriched this dataset with weather variables (e.g. temperature,...
0
votes
1answer
46 views

varying classification threshold to produce ROC curves.

How can I vary classification threshold to produce ROC curves. I am new to R and I wanted to classify in different algorithm. Since the accuracy of ROC plot is high I wanted to change the threshold ...
0
votes
0answers
23 views

Machine Learning: How to solve “class imbalance” in Regression Algorithms?

I have a time-series dataset collected by a group of biologists counting the abundance of a particular animal species in an area. I later enriched this dataset with weather variables (e.g. temperature,...
0
votes
1answer
44 views

Understanding the role of the discount factor in reinforcement learning

I'm teaching myself about reinforcement learning, and trying to understand the concept of discounted reward. So the reward is necessary to tell the system which state-action pairs are good, and which ...
15
votes
3answers
301 views

How to deal with hierarchical / nested data in machine learning

I'll explain my problem with an example. Suppose you want to predict the income of an individual given some attributes: {Age, Gender, Country, Region, City}. You have a training dataset like so <...
3
votes
1answer
30 views

regarding the convolutional network structure of FCNN

The paper of Fully Convolutional Networks for Semantic Segmentation , gives the following image, . What do those numbers represent, 96, 256, 384, etc? Are them ...
-3
votes
0answers
29 views

The proof of lasso regression solution? In which it shrinks some of coefficents to zero? [duplicate]

I would like to know how the lasso method shrinks some of coefficients exactly to zero, "for example could show me how that works mathematically if there are two parameter". For example, with ridge ...
0
votes
0answers
32 views

Observation dependency in logistic regression to learn relevance of search results

I want to predict the most relevant item from a set of search results resulting from a query. Moreover, these items are places; the query is at a lat/long and time; and the search results are ...
1
vote
1answer
46 views

Distribution of the sum of the two dependent bivariate gaussian distributions and related questions

This is something I was thinking about and I decided to modify a question from a mid-term to ask this. Suppose $X_{1}$ and $X_{2}$ are two bivariate gaussian variables, decribed as $$ X_{i}=\begin{...
0
votes
1answer
16 views

regarding computing output size for convolutional layer

I am following up the lecture notes posted on http://cs231n.github.io/convolutional-networks/ I am sort of confusing about one example given in the notes. It says ...
0
votes
1answer
24 views

How do I know if a model with a subset of the features of another model has lowest training/test error?

I'm doing the Machine Learning specialization from the University of Washington on Coursera, and I have to answer some questions in a quiz from the Regression course. They ask which model would have ...
0
votes
0answers
6 views

The derivative of the absolute value |x| [migrated]

I read about the derivative of the absolute value |x|, but why the absolute value is not differentiable at point zero, and when it becomes 1 or -1 {geometrically}? Thanks
1
vote
1answer
35 views

normalisation in k means clustering on percentages and other numerical variables

I have several variables to include in k-means, some of them are percentages (between 0-1) and some of them are numerical variables (positive values). I know normalisation is required when the ...
0
votes
1answer
67 views

Measuring Accuracy of the SVM based model

I have developed a model which evaluates a user based on how important he is for the organization. For that purpose I have generated 1000 records for 1000 users. Here I have one dependent variable "...
3
votes
1answer
28 views

Distribution of classes in neural network batches

I'm creating a neural network for classifying input data. When using batches, do I need to ensure a somewhat uniform distribution of classes per batch? Or can I simply split up my data in any which ...
0
votes
0answers
17 views

bayesglm how to incorporate prior knowledge to improve model

I have downloaded the "University rankings data", available at: https://www.kaggle.com/mylesoneill/world-university-rankings, where the worlds top unis are nked in csv files and I want to take the ...
0
votes
0answers
21 views

Logistic regression with separated data

How can I find the maximum-likelihood estimation of logistic regression parameters when data is completely separated? By completely separated I mean there exist some $x_{0}$ for which classes are ...
0
votes
0answers
20 views

How does the symmetric matrix A= X' X is found to be QDQ'? [duplicate]

I am confused of how does the symmetric matrix A= X'X is found to be QDQ' ? where D is the diagonal matrix of A, and Q is orthogonal matrix i.e Q'Q=I identity matrix.
0
votes
1answer
46 views

Caret Package in R [closed]

I'm starting with the caret package in R and I was wondering when we use train() if all the work was done by the function? I ...
0
votes
0answers
7 views

how principal component analysis find independent class between the data?

i have x=100 with 2D I'ld like to reduce dimension using PCA . my question is how i use 2 class i mean 2 color in my data using matlap.i need full example for reduce dimension using PCA by matlap not ...
1
vote
2answers
47 views

Reduce feature levels

I would like to know if someone knows of a way to group the number of levels of a feature that has 100's (even 1000's) of levels to a smaller number of levels - also, what number levels it should ...
0
votes
0answers
20 views

Minimum sample size for non-stationary time series predictions

I'm not an expert about this topic. I'm trying to make a model to predict cpu usage. Imagine that I want to predict 5 months, What is the minimum sample size to perform the regression?. I've read a ...
6
votes
0answers
59 views

Restricted Boltzmann Machine : how is it used in machine learning?

Background: Yes, Restricted Boltzmann Machine (RBM) CAN be used to initiate the weights of a neural network. Also it CAN be used in a "layer-by-layer" way to build a deep belief network (that is, to ...
1
vote
2answers
37 views

What is the purpose of including negative samples in a training set?

I am new to machine learning and advanced statistics (anything beyond first-year college statistics), and I have been exploring the effectiveness of various classifiers on modeling different hand ...
0
votes
0answers
11 views

Interpretation of `int.size` values from `gbm.interactions` in R

Does anyone know how to interpret the int.size values from the gbm.interactions output? Do they say anything about significance of the interaction? In the table below, can any of these interactions ...
1
vote
0answers
31 views

Which machine learning techniques are applicable to finance? [closed]

I am a mathematician and it is a couple of years since I last implemented any machine learning algorithms. I would like to get back into the field and familiarize myself with the latest research so it ...