Model Architecture 🧠

Technical details about the neural network predicting Collatz sequences

Neural Network Details

Model Architecture

Input Layer 3 features
Dense 128 units (ReLU)
Dropout 20%
Dense 64 units (ReLU)
Dropout 20%
Output 1 unit (Linear)

Training Details

Optimizer

Adam (lr=0.001)

Loss Function

Mean Squared Error

Batch Size

32

Early Stopping

Patience: 20 epochs

Model Performance

11.34
Mean Absolute Error
0.9147
R² Score
10,000
Training Examples
Training History

Input Features

Original Number

The input number to the Collatz sequence

Parity

Whether the number is even (0) or odd (1)

Logarithmic Value

Natural log of (number + 1) for normalization