What causes slow convergence in Keras models?

Slow convergence is caused by poor initialization, low learning rate, or ill-conditioned optimization surfaces.

If gradients are too small or poorly scaled, updates take longer to propagate. Activation saturation, unnormalized inputs, or unsuitable optimizers also contribute.