Model Training

Category: science

The iterative process of teaching an ML algorithm by feeding it large datasets.

Training is the "education phase." You iterate through the data, adjusting the model’s "weights" (importance) until its error rate reaches an acceptable minimum. This process is computationally expensive and requires massive cloud-scale parallel processing.

Common Examples

• The model training process for our predictive pricing engine takes 48 hours to execute on a distributed GPU cluster.
• Every time we update the dataset, we must undergo full model training to ensure the weightings stay accurate to the current market.