otratsiuk/CourseWork

Multilayer perceptron model with backpropagation learning rule

Approximates and forecasts functions given as first order ODE systems.

Network optimization can be achieved by adjusting parameters(alpha, input layer size, hidden layer size and seed for weights and biases random initialization) for a specific function.

---

Example

ẋ = -y - z
ẏ = x + 0,1y
ż = 0,1 + z(x – 5)

parameters for X: alpha = 0.001, E_min = 0.1, input_size = 16, hidden_size = 50, seed = 67
parameters for Y: alpha = 0.001, E_min = 0.1, input_size = 16, hidden_size = 50, seed = 67
parameters for Z: alpha = 0.01, E_min = 0.001, input_size = 16, hidden_size = 50, seed = 67

Approximation	Forecast