You're running an AI project and want to speed up training of the model so that you can complete your current CPMAI iteration within the two week timeframe the team has set. What's one approach to speed up model training?

Extending or fine-tuning a pre-trained model exploits knowledge learned from massive corpora, so only a fraction of data and iterations are needed to reach target accuracy. This transfer-learning approach can cut training times from days to hours, enabling the team to finish the CPMAI iteration within the two-week sprint.

A: Re-using old data may still require full training cycles; speed gains are minimal. B: Cloud offers scalable hardware but does not inherently reduce the number of training epochs. C: Brute-force search increases compute cost and usually lengthens, not shortens, training.

1. Ruder, M., Peters, M. E., Swayamdipta, S., Wolf, T., “Transfer Learning in NLP”, JMLR 2019, §1-2 (pre-trained models reduce training time).

2. MIT 6.S191 “Deep Learning”, 2022, Lecture “Fine-Tuning Pre-trained Models”, slide 7.

3. Cognilytica, “Cognitive Project Management for AI (CPMAI) v7”, 2022, Phase IV best-practice note, p. 68 (use of pre-trained models to accelerate iterations).