Let the machine learn artificial intelligence by itself, "OneClick.ai" is creating an automated deep learning artificial intelligence platform
In recent years, one of the limitations faced by the AI ​​field is that its own technology research and development still requires huge manual investment. From data cleaning, feature engineering, algorithm design, model training and optimization, to the implementation and deployment phases, manpower and time costs are high, which brings great trouble to the landing and popularization of AI.
36 One OneClick.ai has recently developed an automated artificial intelligence platform that can quickly implement the entire process of AI algorithm modeling, tuning, training, implementation and deployment, and automate the use of artificial intelligence applications.
OneClick started research and development in 2016 and received financing support from innovative workshops in 2017. This is also the only artificial intelligence platform that it invests in the United States. At the end of 2017, the OneClick.ai platform was officially launched, supporting deep learning and traditional machine learning algorithms, processing structured data, and unstructured content mixed with text, images, and time series.
In the field of algorithm automation, Microsoft and Google also handed over the answer sheets. In 2017, Microsoft introduced Custom Vision Services, a fully automated platform that supports image classification, allowing users to create CNN models with a small number of images for $29 a month. In 2018, Google also launched a similar Cloud AutoML. Currently, it supports computer vision models. It has not been opened to the public and can be applied for trial.
OneClick.ai founder and CTO Jiang Ning said that solutions like Google AutoML mainly evaluate the design of different neural networks based on search algorithms, and have certain assumptions about the structure of neural networks, so it is only suitable for image classification problems. It's hard to be truly universal.
OneClick.ai is based on the more generalized Generalized Architecture Search (GAS), which can handle classification and regression analysis of more complex data. GAS supports automatic feature engineering and data preprocessing. Like human algorithm engineers, GAS will observe raw data and use supervised learning to accumulate industry knowledge from data to improve computational efficiency.
On the model optimization iteration, Jiang Ning told 36 æ°ª that the platform used a large number of public data sets for training early. After the product is released publicly, every model trained on the OneClick.ai platform will also help the GAS algorithm to improve itself. The Innovation Workshop had conducted a four-month technical assessment before deciding to invest.
OneClick.ai supports the needs of different scenes such as images, text, numbers, and timing. Platform users currently have three ways to use well-trained models: an automated deployment of real-time APIs, offline data processing via chat bots, and an interactive interface for images.
OneClick.AI's user interface allows users to generate predictions with the help of chatbots.
In actual use, OneClick.ai supports cloud and local deployment. With a monthly subscription system, a variety of personal and corporate packages are available for different data processing needs, and users can register for free online.
From the user feedback, OneClick.ai's most popular application scenarios are sales forecasting, inventory and supply chain management. Especially for companies with fast consumption and short shelf life, forecasts can help save inventory, logistics and operations costs, and thus improve business efficiency.
Regarding the predictive efficiency of OneClick.ai, Jiang Ning told 36 that in the consumer industry, their customers are generally companies with sales of more than 100 billion yuan, and a team of more than 100 people is responsible for sales forecasts. In the past, the average accuracy of up to 80% was average, and it was generally aimed at the whole, and the accuracy was lower on specific items. With OneClick.ai you can keep your item sales forecast average accuracy above 90% and the forecasting process takes just a few hours.
In addition to sales forecasting, common application scenarios for this platform include customer retention, advertising recommendation, image classification, object recognition, text classification, and medical image diagnosis. At present, the main customers are large enterprises in the automotive, consumer goods, manufacturing, petrochemical and other industries, with an average size of more than 200 billion yuan.
Talking about business development this year, Jiang Ning said that it will further improve its sales, inventory, and supply chain forecasting capabilities, and form an automation solution that does not require IT intervention. It will continue to improve platform performance and meet customers' artificial intelligence in big data scenarios. demand.
On the team side, OneClick.ai's main R&D team is in Seattle, USA, and its co-founders have all worked at Microsoft. CEO Shen Yuan has extensive experience in AI applications in medical imaging, image search, and e-commerce. CTO's Jiang Ning was responsible for Bing's search advertising relevance algorithm and local search and map search business, with rich experience in algorithm and team management. Shen Yonghui, vice president of sales, is a founding member of the former Tableau China and has 17 years of experience in big data, business intelligence, and artificial intelligence.
Induction Coils,Electronic Eye Induction Coils,Rectangular Induction Coils,Induction Coils For Medical Industry
Shenzhen Sichuangge Magneto-electric Co. , Ltd , https://www.scginductor.com