Data augmentation

Data augmentation is a technique in machine learning used to reduce overfitting when training a machine learning model, by training models on several slightly-modified copies of existing data.

Synthetic oversampling techniques for traditional machine learning

Main article: Oversampling and undersampling in data analysis § Oversampling techniques for classification problems

Data augmentation for image classification

Data augmentation plays a crucial role in image classification tasks, which involve categorizing images into predefined classes. Image classification has numerous applications, including object recognition, scene understanding, and facial recognition.

In image classification, data augmentation helps prevent overfitting and improves the model's generalization capabilities, especially when dealing with limited datasets. By creating slightly modified copies of existing images, data augmentation increases the diversity of the training set, enabling the model to learn from a wider range of examples.

There are several common image augmentation techniques used for image classification tasks:

1. Rotation: Rotating the image by a certain angle, usually within a specific range (e.g., -45 to 45 degrees).
2. Scaling: Resizing the image by a scaling factor, which can either enlarge or reduce the image size.
3. Flipping: Flipping the image horizontally or vertically, creating a mirrored version.
4. Cropping: Removing parts of the image, usually by taking random crops of a predefined size.
5. Brightness and contrast adjustments: Modifying the brightness and contrast levels of the image.
6. Noise injection: Adding random noise to the image, such as Gaussian noise or salt and pepper noise.
7. Color transformations: Changing color properties, such as converting to grayscale, or adjusting hue and saturation levels.

Several popular libraries and frameworks offer image augmentation functionality for image classification tasks:

1. Albumentations: A high-performance, flexible library for image augmentation built on top of OpenCV.
2. Keras ImageDataGenerator: A built-in class in Keras for real-time data augmentation with various image preprocessing functions.
3. imgaug: A versatile library for image augmentation with a wide range of available techniques and configurations.
4. torchvision (for PyTorch): A package containing popular datasets, model architectures, and common image transformations for computer vision in the PyTorch framework.

When applying data augmentation in image classification tasks, it is essential to select appropriate augmentation techniques and their parameters based on the specific problem domain and dataset characteristics. Practical examples and best practices include:

• Evaluating the impact of different augmentation techniques on the model's performance, using techniques such as cross-validation.
• Gradually increasing the complexity and variety of augmentations to prevent the model from overfitting while maintaining its ability to learn meaningful features.
• Ensuring that the chosen augmentation techniques are consistent with the problem domain and do not introduce unrealistic or irrelevant variations to the dataset.

Data augmentation for signal processing

Residual or block bootstrap can be used for time series augmentation.

Biological signals

Synthetic data augmentation is of paramount importance for machine learning classification, particularly for biological data, which tend to be high dimensional and scarce. The applications of robotic control and augmentation in disabled and able-bodied subjects still rely mainly on subject-specific analyses. Data scarcity is notable in signal processing problems such as for Parkinson's Disease Electromyography signals, which are difficult to source - Zanini, et al. noted that it is possible to use a Generative adversarial network (in particular, a DCGAN) to perform style transfer in order to generate synthetic electromyographic signals that corresponded to those exhibited by sufferers of Parkinson's Disease.

The approaches are also important in electroencephalography (brainwaves). Wang, et al. explored the idea of using Deep Convolutional Neural Networks for EEG-Based Emotion Recognition, results show that emotion recognition was improved when data augmentation was used.

A common approach is to generate synthetic signals by re-arranging components of real data. Lotte proposed a method of "Artificial Trial Generation Based on Analogy" where three data examples $x_{1},x_{2},x_{3}$ provide examples and an artificial $x_{synthetic}$ is formed which is to $x_{3}$ what $x_{2}$ is to $x_{1}$. A transformation is applied to $x_{1}$ to make it more similar to $x_{2}$, the same transformation is then applied to $x_{3}$ which generates $x_{synthetic}$. This approach was shown to improve performance of a Linear Discriminant Analysis classifier on three different datasets.

Current research shows great impact can be derived from relatively simple techniques. For example, Freer observed that introducing noise into gathered data to form additional data points improved the learning ability of several models which otherwise performed relatively poorly. Tsinganos et al. studied the approaches of magnitude warping, wavelet decomposition, and synthetic surface EMG models (generative approaches) for hand gesture recognition, finding classification performance increases of up to +16% when augmented data was introduced during training. More recently, data augmentation studies have begun to focus on the field of deep learning, more specifically on the ability of generative models to create artificial data which is then introduced during the classification model training process. In 2018, Luo et al. observed that useful EEG signal data could be generated by Conditional Wasserstein Generative Adversarial Networks (GANs) which was then introduced to the training set in a classical train-test learning framework. The authors found classification performance was improved when such techniques were introduced.

Mechanical signals

The prediction of mechanical signals based on data augmentation brings a new generation of technological innovations, such as new energy dispatch, 5G communication field, and robotics control engineering. In 2022, Yang et al. integrate constraints, optimization and control into a deep network framework based on data augmentation and data pruning with spatio-temporal data correlation, and improve the interpretability, safety and controllability of deep learning in real industrial projects through explicit mathematical programming equations and analytical solutions.

See also

• Oversampling and undersampling in data analysis
• Generative adversarial network
• Variational autoencoder
• Data pre-processing
• Convolutional neural network
• Regularization (mathematics)
• Data preparation
• Data fusion