Functional data learning using convolutional neural networks

Jose Galarza,Tamer Oraby

2023-10-05

Abstract:In this paper, we show how convolutional neural networks (CNN) can be used in regression and classification learning problems of noisy and non-noisy functional data. The main idea is to transform the functional data into a 28 by 28 image. We use a specific but typical architecture of a convolutional neural network to perform all the regression exercises of parameter estimation and functional form classification. First, we use some functional case studies of functional data with and without random noise to showcase the strength of the new method. In particular, we use it to estimate exponential growth and decay rates, the bandwidths of sine and cosine functions, and the magnitudes and widths of curve peaks. We also use it to classify the monotonicity and curvatures of functional data, algebraic versus exponential growth, and the number of peaks of functional data. Second, we apply the same convolutional neural networks to Lyapunov exponent estimation in noisy and non-noisy chaotic data, in estimating rates of disease transmission from epidemic curves, and in detecting the similarity of drug dissolution profiles. Finally, we apply the method to real-life data to detect Parkinson's disease patients in a classification problem. The method, although simple, shows high accuracy and is promising for future use in engineering and medical applications.

Machine Learning,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper primarily explores how to use Convolutional Neural Networks (CNN) to address learning problems of Functional Data (FD), including regression and classification tasks. Specifically, the paper attempts to solve the following key issues:

1. **Transforming Functional Data into Images**: The paper proposes a method to convert functional data into 28x28 pixel grayscale images to facilitate processing with CNN.

2. **Parameter Estimation**: Using CNN to estimate parameters of different functions, such as the growth or decay rate of exponential functions, the frequency of sine and cosine functions, the height and width of curve peaks, etc.

3. **Functional Data Classification**: Using CNN to classify functional data, including distinguishing the monotonicity of functions (increasing vs. decreasing), curvature (convex vs. concave), growth type (exponential vs. algebraic), and identifying the number of peaks in curves.

4. **Application to Real-world Problems**:
- **Dynamic System Analysis**: Using CNN to estimate the Lyapunov exponent of the Lorenz system, which is an indicator of system stability.
- **Infectious Disease Transmission Rate Estimation**: Estimating transmission rates through simulated epidemic curves, which is crucial for understanding the development of pandemics.
- **Drug Dissolution Curve Similarity Detection**: Using CNN to compare the dissolution curves of different drugs to evaluate their similarity.

5. **Practical Case Study**: Finally, the paper applies the same method to classify real cases of Parkinson's disease patients, validating the effectiveness and accuracy of the method.

In summary, the core objective of this paper is to demonstrate how to effectively use CNN to solve various problems in functional data learning and to prove that this method has high accuracy and potential value in engineering and medical applications.