Questions tagged [data-augmentation]

Data augmentation is the practice of making slight modifications to the observed data with the goal of making models trained on that data more robust.

Filter by
Sorted by
Tagged with
1 vote
0 answers
20 views

Decide on the number of adversarial samples to include during training

A sample is considered adversarial if it drastically changes classifier's confidence $\in [0,1]$ when given as input. For example, if a spam/ham binary classifier considers some input $X$ to be 0.9 ...
user avatar
  • 224
2 votes
1 answer
142 views

Does it make sense to use data augmentation on the Validation set? (note, this is not the same as asking to augment the test set)

Curious, do people use data augmentation on the validation set? I am aware there is a debate for the test set -- but the validation set is usually a split form the train set, so wouldn't it make sense ...
user avatar
1 vote
2 answers
81 views

How to perform data augmentation with traditional machine learning algorithms?

I am currently working on a multi-class image classification project, in which I have to use traditional machine learning and feature extraction methods (no convolutional neural networks). I know data ...
user avatar
  • 11
0 votes
1 answer
55 views

How do we perform hyperparameter tuning on parameter of data augmentation?

I was wondering how do we perform hyperparameter tuning on parameters of data augmentation. Suppose I have to select the best pair of (alpha, alpha) values of beta distribution that works best on data ...
user avatar
  • 13
1 vote
0 answers
34 views

Is it "cheating" to augment data by adding noise to the label?

I have a data set that I'd like to augment by adding noise to the label. I've seen people on Kaggle duplicate each row twice and add 1 and subtract 1 to the label Whenever I do this and then do a ...
user avatar
  • 11
0 votes
0 answers
30 views

Why my validation accuracy is higher than my train accuracy after augmentation?

I'm using a text dataset of 10158 instances where I convert each instance into sentence embedding before training also I'm using different augmentation techniques but there is this one specific ...
user avatar
  • 13
0 votes
0 answers
20 views

Mixup VS CutMix Data Augumentation Method

I am looking for arguments on which Data augmentation (Mixup VS CutMix) method would be preferable for Image data and Time-series classification data. As for as I know, both have the following ...
user avatar
  • 121
0 votes
0 answers
23 views

ROSE acceptable dispersion/shrinkage

To solve imbalanced data, I used oversampling strategy using ROSE algorithm in Python. As you may know, ROSE is a smoothed bootstrapping method and we can control the dispersion of the augmented data. ...
user avatar
2 votes
1 answer
96 views

Does data augmentation with white noise improve accuracy of deep learning models?

I was reading Aurélien Géron's Hands-on Machine Learning with Scikit-Learn, Keras & TensorFlow. There, on the 14th chapter I read something on data augmentation which I could not be sure of its ...
user avatar
0 votes
0 answers
43 views

Validation loss not decreasing & Poor performance on a Regression task

Outline Hello, I am training a DNN to predict the concentration (an unbound value) of Tryptophan (an amino acid) in water. The data is 1D Raman spectra. I have tried several different architecture ...
user avatar
  • 101
0 votes
0 answers
18 views

What is the equivalent of image augmentation in time series forecasting? I'm in need for more data [duplicate]

I think it would be fair for me to explain a little bit on a background into what I am doing so that my question would make more sense. I am currently working at a company where I need to develop a ...
user avatar
1 vote
0 answers
29 views

Overfitting, generalization, data augmentation, regularization, how do they relate to each other? How to measure?

Recent work such as Deep Double Descent shows that overfitting is not really a problem with large models, even without any data augmentation or regularization (L2 weight norm, dropout or so). Edit: Ok,...
user avatar
  • 1,155
0 votes
0 answers
54 views

Using data augmentation for balancing dataset

Hey there, I have a question about a topic that's been discussed many times before, but to which I could find a satisfying answer. I'm working with a self generated dataset that is comprised of only ...
user avatar
1 vote
1 answer
29 views

When augmenting data, shall the dataset keep a balanced ratio

When training a model it is more and more common to augment data. posts indicate that only the training set shall be augmented. On the other hand it is common to split dataset in a fashion following ...
user avatar
  • 11
0 votes
0 answers
71 views

Time Series: Do I understand Windows Slicing correctly?

So in the following Thread it is discussed about augmentation for time series: Data Augmentation strategies for Time Series Forecasting The first answer refers among others to the following: Window ...
user avatar
0 votes
0 answers
93 views

Based on Data Augmentation in numerical dataset

I want to develop a machine learning model for a dataset like 5 inputs and 1 output. I stuck in stage with not sufficient dataset, I am aware about data augmentation technique in image and text type ...
user avatar
  • 1
1 vote
1 answer
101 views

Conditional distribution of the weight of a mixture gaussian with data augmentation using gibbs sampling

This question is relate to Differenciate between two distributions using gibbs sampling . For $t=1,\,\dots,\,n$, let's $r_t\sim\mathcal{N}(0,\,\sigma_t^2)$ and $$\sigma_t^2=\left\{\begin{array}{lcl} \...
user avatar
0 votes
0 answers
20 views

Forcing uniform prior when training classifier?

Say you're training a classifier to take an input $x$ and predict its label $y \in \{1,\ldots,k\}$. As an example, let's say the classifier is a neural net, which ends in a softmax layer, and we train ...
user avatar
  • 379
0 votes
0 answers
24 views

Validation Loss gets better after adding Augmentation Layers but Test accuracy gets worse

I'm building a Siamese Network which should learn a face comparison function. My model consists a CNN (which gets 2 inputs, and yields 2 embedding vectors). With the outputs I calculate: ...
user avatar
1 vote
1 answer
167 views

Why Accuracy increase only 1% after data augmentation NLP?

i have small dataset 4840 samples (60% negative ,28% positive,12% negative) i use data augmentation on training set (70%train 30% test) and i have about 2000 samples for each class while test is ...
user avatar
1 vote
0 answers
37 views

From 1 to 5-shot learning with data augmentation

I'm currently exploring k-shot classification and I would like to start a first experiment on logo classification. The problem is that for some logos I can only find one image, while for others I ...
user avatar
  • 11
0 votes
0 answers
27 views

How to deal with data augmentation for training neural networks?

I'm trying to apply some data augmentation techniques for training my neural network model. I know that I need to avoid including synthetic data generated from test data in the training data. Also, I'...
user avatar
  • 378
0 votes
0 answers
99 views

Mask for image padding in semantic segmentation

I'm using data augmentation for a semantic segmentation task, where some images are cropped or rotated. As a result, some padding is added to ensure that the image is always the same size. These ...
user avatar
6 votes
1 answer
187 views

Does oversampling lead to more overfitting than classweights for really small classes?

Assume I have a couple of thousand hens that I want to classify into those that never lay an egg and those that will at some point in their life lay an egg. Assume that already works perfectly. Now ...
user avatar
0 votes
0 answers
33 views

The labels using MixUp data augmentation in Kernel SVM (dual form)

As we know that mixup data augmentation do convex combination on the label $\lambda y_i + (1-\lambda)y_j$, $\lambda\sim Beta(\alpha,\alpha)$, assume that in binary classification, our label can only ...
user avatar
0 votes
0 answers
57 views

Data augmentation for traditional machine learning algorithms

Data augmentation suffices multiple purposes, I would list a few here: Increasing dataset size: The data is just fragment/stand-in trying to represent reality, having more data should thus result in ...
user avatar
  • 140
1 vote
0 answers
16 views

Learning Distribution of Data [duplicate]

Sometimes it's important to generate data due to data imbalance issues. I heard that data augmentation by leaning distribution of data is a hot topic now. Could you please give me some resources and ...
user avatar
  • 113
0 votes
2 answers
54 views

How to generate synthetic data from a balanced dataset?

Let say I have a balanced dataset that has a small training sample size (lack of data). How do I increase the training sample size by generating synthetic data based on the original data? I believe ...
user avatar
  • 9
1 vote
1 answer
65 views

What is the best practice to overcome small insufficient data

I have a small number of images (i.e. 108), and I wish to train a deep convolutional neural network with it. As I know - you need to have a large number of samples to be able to train a neural network,...
user avatar
5 votes
1 answer
73 views

In a parametric model, if I do not have enough data, can I estimate the parameter, and simulate data from the estimated model and estimate again?

Suppose I have a logistic regression model $Y_i=\mathbf{1}(X_i\beta>\epsilon_i)$ to estimate, where the distribution of $\epsilon_i$ is known, $X_i$ follows distribution $F_{\theta}$ with an ...
user avatar
  • 1,662
3 votes
1 answer
48 views

Expert Knowledge Acquisition and Machine learning

Having data sets regarding symptoms and diseases such that I use to observe the conditional distributions P(Disease X | Symptom A , Symptom H , Age >20 ) which I use for classification and ...
user avatar
  • 320
0 votes
0 answers
993 views

Data augmentation by adding noise in python regression model

I am building a regression model for a target variable which is heavy tailed. I want to augment data so that the model gets enough training samples in the region where it's a long tail. Accuracy of ...
user avatar
1 vote
1 answer
186 views

Train test validation splits and augmentation

I am dealing with an image dataset of 400, and split it into 70% train, 15%test, 15%validation. I would like to do some data augmentation (rotations/flips) to increase the amount of train data I have ...
user avatar
1 vote
0 answers
53 views

Data Augmentation causing test and validation sets to be smaller

I am dealing with an image dataset of 400, and split it into 70% train, 15%test, 15%validation. I would like to do some data augmentation (rotations/flips) to increase the amount of train data I have ...
user avatar
2 votes
1 answer
20 views

Augmenting training data with cases that won't be in future data

Background: I am working on coding survey responses, where the respondent writes in a description of their job. So the person might write in "McDonald's Employee" and get coded to something like 1002 ...
user avatar
  • 1,408
1 vote
0 answers
131 views

Test Time Augmentation on Validation set?

In the traditional usage of data augmentations, we augment only the train set examples, in order to keep the distribution of the validation and test set equal. In the TTA method, we apply ...
user avatar
  • 201
0 votes
0 answers
140 views

How to train a neural network with an incomplete dataset?

I am currently training a neural network with a dataset containing approximately 10 features and 1000 entries. The problem is that 70% of the entries contain at least one missing value for at least ...
user avatar
  • 121
20 votes
3 answers
2k views

What are the mathematically rigorous data augmentation techniques?

Imagine you have a dataset of 1000 observations. To keep things intuitive imagine they are (x,y) coordinates. They are temporary independent, so that makes it easier. You wish you had about a million ...
user avatar
  • 453
2 votes
0 answers
107 views

MCMC converges to MAP and stays at same value - what may go wrong?

I am working on a Gibbs sampler which is complex and I would like to avoid giving all the details here. I will focus on the most necessary details. The Gibbs sampler involves parameters and latent ...
user avatar
  • 5,934
6 votes
0 answers
116 views

Does EM algorithm require us to know the joint (predictive) distribution of the latent variables $Z$ when $Z$ is two-dimensional?

In its general form the E-step of the EM algorithm finds the expectation $$ Q(\theta|\theta') =\int \log[ p(Y,Z | \theta)] p(Z|Y,\theta') d Z$$ where $Y$ the data, $Z$ the latent variables, $\theta'$...
user avatar
  • 5,934
0 votes
0 answers
121 views

Generate more data for a small dataset

I have been working on a dataset which has 14 attributes and 303 rows(instances) along with the binary labels. I want to generate more data so that I could train my neural networks so that I could ...
user avatar
0 votes
1 answer
490 views

is this way of applying data augmentation correct [closed]

I'm training a CNN and want to apply some data augmentation to my input images. I combined some code from tensorflow tutorials and have the following workflow: I have a dataset containing all ...
user avatar
8 votes
1 answer
134 views

Structure of Generative Adversarial Networks (GAN) for mapping a simulation model

There is a simulation model of a system that I want to map as a neural network to test if a better execution time can be achieved with similar accuracy. The simulation model receives real-valued ...
user avatar
  • 93
0 votes
0 answers
134 views

Is GAN effective enough to replace data augmentation and manual annotation?

We all know that GAN can be used to augment and expand our dataset Can a GAN be used for data augmentation?. But my question is, is it effective and fast enough? For example I have done experiment ...
user avatar
  • 161
2 votes
0 answers
264 views

Can GANs be used for timeseries data augmentation? (2019)

Timeseries, in particular signal timeseries, are distinct in many respects - so GANs working on images may not work for timeseries. Since other questions asking on data augmentation, GANs have ...
user avatar
4 votes
0 answers
1k views

GANs for non image data

I'm looking to narrow down the subject for my bachelor thesis: I am currently working on a project, that only offers a small dataset and there will be no more data incoming for now. What I'm trying to ...
user avatar
1 vote
0 answers
14 views

Variation in accuracy of data splitting before and after data augmentation

How much accuracy of the system varied/changes between two cases Data augmentation before splitting Data augmentation after splitting, only on training data Is there any literature published?
user avatar
  • 151
1 vote
1 answer
569 views

Data augmentation on entire dataset before splitting

If I apply rotation of 5 different angles and randomly cropp 10 different images from each rotated image and than divided into training testing and validation. Will it be totally incorrect evaluation ...
user avatar
1 vote
0 answers
1k views

Data augmentation techniques for numeric datasets? [duplicate]

I'm writing a paper about Data Augmentation and I'm looking for some way of increasing the size of a dataset. I'm already aware of the techniques used for images (transformations, PCA, blurs, etc.) ...
user avatar
  • 11
1 vote
1 answer
315 views

Why is using keras ImageDataGenerator for data augmentation relevent?

I have used keras ImageDataGenerator to generate more data in my neural networks as I have had really small datasets and it has proven itself. As far as I ...
user avatar
  • 189