Amazon Web Services, DeepLearning.ai, and Coursera seek to bridge the gap between machine learning and model building and testing, and their scaling in production through three-course specialization.
“Machine learning has a proof of concept for production gaps,” explains Andrew Ng, founder of DeepLearning.AI and top instructor at Coursera. This specialization is designed to allow developers to move models from prototypes on laptops to the cloud. “There’s a lot to do to move from 10 users to 1 million users,” Ng added.
According to Bratin Saha, vice president and general manager of machine learning services at Amazon AI, AWS customers have deployed from a few models to millions of models in just a few years. “ML is no longer a niche,” said Saha, who oversees SageMaker, AWS’s fastest-growing product, the machine learning platform.
The specialized course provides an overview of the moving parts (MLOps, DevOps) required to move your model to production and topics on accuracy, cost, and optimization as prototype scaling.
Primer: What is AI? | What is machine learning? | What is deep learning? | What is general-purpose artificial intelligence?
In an interview with Ng and Saha, I touched on some notable points about the model. Highlights of our chat:
Does the model need to start in the cloud from the beginning, given the scale? Ng’s approach to machine learning models is based on “using the right tools for the right job.” “It’s okay to do a proof of concept on a laptop. To do a proof of concept, you have to decide whether to go or not,” says Ng.
Ng added that planning the scale before the proof of concept can disrupt the process.
Skills required for scalingAccording to Saha, specialized courses are designed to expand the talent base of machine learning. Both Saha and Ng say they lack the talent to understand how to scale their models. “Two years ago we trained our models with 20 million parameters, but now we have 100 million. We make hundreds of billions of forecasts a month,” says Saha.
Ng said the demand for skilled machine learning practitioners is high, and there is an even shortage of people deploying meaningful services in the cloud. As a result, Saha said that all engineers joining Amazon are taking mandatory machine learning courses.
Machine learning is in the early stages of its evolution. Ng said that in many ways machine learning rhymes with early software development. “I vaguely remember when software engineering was in turmoil, but now version control is even more mature,” says Ng. “I’m inspired by how software emerged as an industry.”
Also: Andrew Ng foresaw the eternal spring of AI
Here are the key points for Practical Data Science disciplines.
- This discipline is familiar with the Python and SQL programming languages and is designed for data-centric developers, scientists, and analysts who want to build an end-to-end machine learning pipeline.
- Algorithms for natural language processing and natural language understanding (BERT, GLoVe, ELMo, FastText, etc.).
- The first course covers basic concepts and exploratory data analysis using Amazon SageMaker Studio and other SageMaker services. Learn about automatic machine learning.
- In the second course, learners build, train, and deploy an end-to-end machine learning pipeline.
- The third course covers advanced model training, tuning, and deployment techniques. Learn about distributed training, hyperparameter tuning, and A / B testing.
- A managed online lab environment is provided by AWS partner Vocareum.
Watch Now on Coursera
AWS, DeepLearning.ai aims to close the scaling gap with machine learning models through Coursera specialization
Source link AWS, DeepLearning.ai aims to close the scaling gap with machine learning models through Coursera specialization
Credit: Google News