Neural Differential Equations for Time Series Analysis
At this repository you can find the code that implements the Neural Differential Equation algorithm. The code is commented in detail. Also you can check my report on Neural ODEs: "Understanding the Neural ODEs.pdf", where I describe the idea, intuition and math behind the algorithm, followed by a thorough
breakdown of its implementation.
From ResNet to Neural ODE
Representing an ODE in form of Neural Network Layer
Learning the dynamics of Time Series Data
Forward pass and Backward pass(with Adjoint Sensitivity method)
Visualizing the training and results
$\frac{dz}{dt} = \begin{bmatrix} -0.1 & 2.0 \\ -2.0 & -0.1 \\ \end{bmatrix} z^3 \quad \text{with} \quad \begin{bmatrix}z_1\\z_2\end{bmatrix}(0) \begin{bmatrix}0.0\\2.0\end{bmatrix}$
GIF
$\frac{dz}{dt} = \begin{bmatrix}-0.1 & -1.0\\1.0 & -0.1\end{bmatrix} z \quad \text{with} \quad \begin{bmatrix}z_1\\z_2\end{bmatrix}(0) \begin{bmatrix}0.6\\0.3\end{bmatrix}$
About
Understanding the idea, intuition and implementation of Neural Differential Equations. Clearly explained and fully commented.