Cyclical Learning Rates (CLR) é uma técnica de treinamento em aprendizado profundo that adjusts the taxa de aprendizado dynamically during the training process. Instead of using a taxa de aprendizado constante, CLR allows the learning rate to fluctuate between a lower and upper bound, creating a cycle that can melhorar a eficiência do treinamento do modelo e desempenho.
The concept is based on the idea that periodically increasing the learning rate can help the processo de otimização escape local minima, while decreasing it can facilitate convergence. Typically, during each cycle, the learning rate starts at a minimum value, increases to a maximum value, and then returns to the minimum. This cyclical pattern can be repeated multiple times throughout the training process.
O CLR possui várias vantagens, incluindo:
- Convergência mais rápida: By allowing the learning rate to vary, models can converge more quickly than with a fixed learning rate.
- Desempenho aprimorado: The technique can help models achieve better generalization ao prevenir o overfitting.
- Menor necessidade de ajuste: Since CLR adapts the learning rate during training, it can reduce the need for extensive ajuste de hiperparâmetros.
Common implementations of CLR include triangular cycles, where the learning rate increases and decreases in a triangular waveform, and exponential cycles, where the learning rate follows an decaimento exponencial pattern at certain points. Overall, Cyclical Learning Rates represent a powerful method for optimizing the training of deep learning models, making them a popular choice among practitioners.