NEWS

－Total Support from “Development” to “Management, Operation and Relearning” of AI and Machine Learning Model－

SIGNATE Inc. (Head Office: Chiyoda Ward, Tokyo; President and CEO: Shigeru Saito), a provider of AI development and data scientist human resource recruitment and development services, has released the “Aldebaran” platform for supporting the management, operation and relearning of AI and machine learning models. In addition to the development of AI and machine learning models that SIGNATE previously provided, this platform will enable SIGNATE to provide total support encompassing the management, operation and relearning of the developed models.

■Outline of Aldebaran
The Aldebaran platform supports the management, operation and relearning of AI and machine learning models and is provided as AIaaS (AI as a Service).

Based on support provided by SIGNATE data scientists, companies using Aldebaran can use AI and machine learning models via API simply by configuring initial settings, instead of needing to build an execution environment for such models. The API provided by Aldebaran enables the company to easily connect inhouse applications to the AI and machine learning models.
An optional service is available for customers who do not have inhouse applications whereby SIGNATE also develops a dashboard for viewing the prediction and classification results of the AI and machine learning models, and which enables the entry of operation results data.

SIGNATE data scientists also provide support for relearning in accordance with the AI and machine learning models. In contrast to regular software, the accuracy of AI improves gradually after the start of operations, increasing its value. This characteristic of AI is fully utilized to make customer tasks more efficient and improve service levels.

SIGNATE will continue to develop Aldebaran in order to achieve full automation of processes such as accuracy evaluation, relearning and enhancement learning for the AI and machine learning models.

▼Aldebaran Configuration ​
 

■Aldebaran Functions
(1) Provision of an execution environment for AI and machine learning models
(2) Management of API information
(3) Management of learning data
(4) Entry of operation results data
(5) Relearning support for AI and machine learning models
(6) Version management for AI and machine learning models

■Aldebaran Features
(1) Building an execution environment for AI and machine learning models is unnecessary
(2) Easy to implement AI and machine learning models for inhouse applications via API
(3) Relearning support for AI and machine learning models is also provided

■Used for Effectiveness Verification with Aim of Implementing JR West
Estimation of Snow Accretion to Hokuriku Shinkansen Bogie” Model Aldebaran has already been used for effectiveness verification with the aim of implementing an “AI for Estimating Snow Accretion” to the bogie part of the Hokuriku Shinkansen in the West Japan Railway Company (Head Office: Osaka; President, Representative Director and Executive Officer: Tatsuo Kijima; hereinafter “JR West”). Last year, JR West launched a challenge to estimate the amount of snow accretion to the bogie part of Hokuriku Shinkansen trains while in operation（https://signate.jp/competitions/58）via “SIGNATE,” AI development competition website. Three winning AI and machine learning models were run on Aldebaran to verify the effectiveness of the models, with the aim of implementing the models in the business. The results were verified via the optional service dashboard by comparing current operations to the amount of snow accretion to the bogie part and the necessity of snow removal that were predicted by the three AI and machine learning models. From these results, it was determined whether the prediction model could be used in actual operations.

▼Dashboard of AI for Estimating Snow Accretion​
 

■Total Support from “Development” to “Management, Operation and Relearning” of AI and Machine Learning Models
The dramatic take-off in the utilization of AI and big data across society has led to a global shortage of the AI engineers and data scientists who are most involved in these technologies, and supply cannot keep up with demand. SIGNATE addresses this demand by connecting global talent to companies that have specific AI development or data analysis issues, providing high-precision AI and machine learning models through a competition-based system.
This newly released service enables companies to connect their AI and machine learning models, including those procured via the competition, to the applications used for their regular tasks. SIGNATE provides total support for AI and machine learning models, from “development” to “management, operation and relearning,” and will continue to promote further corporate utilization of AI and big data.