tensorflow - tensorflow 中训练-评估-预测的每个步骤的目的是什么？

Question

每个阶段的作用是什么？据我所知，对于 nlp 中的神经网络，训练会找到单词嵌入的最佳参数。但评估步骤的目的是什么？它应该做什么？这与预测阶段有何不同？

Answer 1

训练、评估和预测是训练模型(基本上在任何机器学习框架中)以及将模型从研究/开发转移到生产的三个主要步骤。

培训:

根据需要解决的问题选择合适的机器学习架构。进行超参数优化来微调模型。然后模型会根据一定数量的 epoch 的数据进行训练。 监控损失、准确性、MSE 等指标。

评价:

We need to move the model to production. The model in the production stage will only make inferences and hence we require the best model possible. So, in order to evaluate or test the model based on some predefined levels, the evaluation phase is carried out.

评估主要是对原始数据集的子集数据进行的。在预处理数据时进行训练和评估分割。 计算指标是为了检查模型在评估数据集上的性能。

The evaluation data has been never seen by the model as it is not trained on it. Hence, the model's best performance is expected here.

预测:

模型测试完成后，我们可以将其投入生产。在生产阶段，模型仅对提供给它们的数据进行推断(预测)。这里不进行任何培训。

Even after a thorough examination, the model tends to make mispredictions. Hence, in the production stage, we can receive interactive feedback from the users about the performance of the model.

现在，

But what is the purpose of the evaluation step? What is it supposed to do? How is that different from the prediction phase?

评估是为了使模型在遇到的大多数情况下变得更好。进行预测是为了检查与性能无关的其他问题。