[Service Development Team Im Changdae]
'Vertex AI' was unveiled at Google's developer conference IO in May. Vertex AI is Google Cloud's managed cloud service and is a machine learning-as-a-service (MLaaS) platform that integrates AutoML and AI Platform.
In addition to Google Cloud, the cloud service providers that provide such a platform are currently competing with SageMaker of AWS and Azure ML Studio of Microsoft Azure. It is attracting the attention of data scientists by claiming that it can reduce the amount of to 80%.
Cloud service providers offer tools such as predictive analytics and deep learning, data visualization, natural language processing, MLOps API, and more. AutoML lets you train models on image, tabular, text, and video datasets without writing any code, and lets you run custom training code with training on AI Platform.
In addition, in terms of computation, it helps to speed up machine learning model training using GPUs and TPUs provided by cloud service providers' data centers. This allows customers to increase the cost and speed of developing ML services and easily use the best-in-class technology.
In addition, Vertex AI includes the following services.
- Data preparation processcollect data from Big Query and cloud storage for Vertex Data Labelscan be used to annotate high-quality training data and improve prediction accuracy.
- Vertex Metadataprovides data lineage and execution tracing capabilities, including artifacts, data origins, and where they occurred over time and where they occurred, through the Python SDK, describing the output generated by the model training process.
- Vertex feature(Feature) allows you to add and select features used by ML, and provides a feature repository for sharing and reuse.
- vertex trainingprovides a set of pre-built algorithms and allows users to import custom code into their training models. You can perform model training using hybrid artificial intelligence, such as on-premise environments or other cloud environments that require more customization.
- Vertex Neural Network Architecture Searchbuilds new model neural network architectures appropriate to service-specific requirements and discovers new structures by optimizing existing neural network architectures for latency, memory, and power.
- Vertex Explainable Artificial Intelligencetells how much each feature of the data contributed to the prediction outcome. You can use this information to verify that the model is behaving as expected, recognize biases in the model, and get ideas for improving the model and training data.
In Korea, numerous domestic companies such as Samsung Electronics, Homeplus, and Kia Motors are currently collaborating with Google Cloud. In particular, Samsung Electronics applied Google Cloud's Big Query, Cloud Spanner, and Dataflow services to its voice recognition platform Bixby to collect data. You are analyzing the data. Vertex AI, which integrates these services, is currently available on Google Cloud Platform, and detailed guides for services such as AI platform and Auto ML are provided.
From building an artificial intelligence model to actual operation, faster and more optimized, it would be good to respond more agilely to the dynamically changing market.
References
Google Cloud Platform Vertex AI: https://cloud.google.com/vertex-ai
