Leifeng.com learned that the artificial intelligence company “Zhiyun”, an artificial intelligence company focusing on small-sample learning, has recently received tens of millions of yuan in Pre-A round financing. Zhepu (Shanghai) Fund, a national small and medium-sized enterprise development fund managed by Zheshang Venture Capital, Leading the investment, this round of financing will be used to strengthen the research and development of X-Brain AIoT technology and iteratively upgrade the AI platform.
Zhongke Zhiyun was established in 2018. The company currently has more than 100 team members. The core team comes from Chinese Academy of Sciences, Oxford University and other institutions, of which nearly 70% are technical R&D personnel. Its main business is centered on the self-developed X-Brain AI active learning platform, which integrates small sample learning framework, multi-source fusion perception computing and other technologies to provide AI security governance services for the industry.
At present, in order to avoid overfitting or underfitting, AI with deep learning as its core needs to use a large amount of data for model training, so that the model can achieve better fit and solve scene problems. However, when faced with fragmented scenarios, this path is difficult to implement due to the lack of data and poor training effect.
For this reason, few-shot learning (FSL) and AutoML (automatic machine learning) are gradually becoming new algorithm production modes.
Compared with traditional machine learning, the advantage of small-sample learning is that it directly trains the algorithm model with a smaller amount of data or samples. AutoML can realize automation in four aspects: feature engineering, model construction, hyperparameter selection, and optimization methods, which not only reduces algorithm production costs, but also improves efficiency, and lowers the threshold for algorithm production.
In an interview with Leifeng.com, Wei Hongfeng, CEO of Zhongke Zhiyun, said that small samples are the basis of low cost, because the sample size is small, and the training model does not require high computing power hardware equipment. However, in some scenarios, the accuracy of small sample training is difficult to reach the commercial level in the early stage. Therefore, AutoML can be used to shorten the process from initial model to commercial use.
The core of the X-Brain platform launched by Zhiyun is a set of active learning algorithm framework, and the self-developed active learning (Active Learning) technology is applied to change the mode of passively accepting manually labeled samples in supervised learning.
The platform can use AI to actively judge whether a sample requires the participation of an algorithm engineer. By only allowing the algorithm engineer to participate in the confirmation of some difficult samples, a human-in-the-loop (Human-in-the-Loop) mode is formed, and the model is actively trained to form an automatic iteration of the model. , so as to solve the problem of high cost of algorithm production.
Up to now, the X-Brain AI platform has been applied in industries such as industry, construction, and electric power.
Wei Hongfeng believes that small sample learning and AutoML technically solve the problem of algorithm production, but how to make the technology better serve the enterprise and solve the problem of the actual scene requires dismantling the pain points of the specific scene and integrating them into the entire algorithm production process. From business and problem definition, to model tuning, and finally to algorithm delivery, people who understand the business must be involved. Leifeng.com Leifeng.com
This article is transferred from: https://www.leiphone.com/category/aijuejinzhi/bjTDaGNvbSMf5UxV.html
This site is only for collection, and the copyright belongs to the original author.