Transwarp Data Hub
Transwarp Data Cloud
Transwarp ArgoDB
Transwarp KunDB
Transwarp Data Studio
Transwarp Sophon
Transwarp Knowledge Hub
TxData (ARM)
TxData (x86)
TxIRA
Selected Product Solutions
Federated Learning
Privacy Protection
/
Connecting Isolated Data Islands
/
Horizontal Federated Learning
/
Vertical Federated Learning
Federated learning
Customized Federated Task Roles
Sophon P²C enables users to select for modeling of a horizontal/vertical federation based on their own scenario data, and customizes the initiators, participants and coordinators in federated learning to complete the role configuration in the task, so that users can make the most efficient use of their data assets and establish a federated model with good performance.
Wizard-style Configuration of Federated Tasks
Sophon P²C provides wizard-style configuration service of the federated task, which supports each participant of the federated task to complete data preprocessing, feature engineering, model parameter configuration and other processes of the federated task according to the Task Wizard. It can quickly complete the task configuration without writing code, effectively reducing the user's learning cost.
Supporting Cryptographic Algorithms and Protocols
Sophon P²C supports homomorphic encryption, differential privacy, oblivious transfer and other cryptographic protocol to protect data exchange and operation process of all parties and prevent data leakage and data reverse engineering. At the same time, it supports secure data transmission based on secure channel, which can effectively ensure the security of network communication.
Federated and Local Feature Engineering
The Sophon P²C platform has abundant built-in federated feature engineering algorithms, including federated missing value filling, federated string index, federated standardization and federated normalization .It can jointly perform data cleaning and feature engineering with data from all parties in some cases where the data is always stored locally and data security is guaranteed, so as to provide data standard guarantee for subsequent federated learning.
Federated Learning Modeling and Evaluation
Sophon P²C supports horizontal and vertical logistic regression, linear regression, K-means and other general algorithms, and also supports the application of a variety of personalized federated learning algorithms in vertical federation scenarios. Moreover it supports the output of multiple indicators for evaluating the models to help users choose the best model and achieve the efficient implementation of AI.
Solution Highlights
Authoritative
Data Security Certification
Sophon P²C is one of the first federated learning platforms in China to pass the qualification of the CAICT, and passes security special assessment and certification in the "Zhuoxin Big Data Plan" of the CAICT. It supports AI cooperation among multiple parties on the premise of data privacy protection. Meanwhile, its R&D team deeply involved in the writing of federated learning industry standards. In addition, it fully understood the industry scenarios and used authoritatively certified platform services to meet the specific needs of users while effectively ensuring users' data security.
Diversified
computing framework
Sophon P²C supports multiple scenarios such as horizontal federated learning, vertical federated learning and federated migration. It also supports self-developed deep learning frameworks such as Sophon, Spark, Tensorflow and Pytorch, and provides expert-level application services, which meet the AI needs of enterprises in a variety of vertical application scenarios and effectively reduce the cost of enterprise decision-making.
Personalized
Federated Learning Modeling
Sophon P²C can complete joint modeling when the data of all participants are not interconnected, so as to realize modeling with data immobility in a true sense and avoid data leakage. At the same time, its personalized federated learning modeling service can fine tune local data and federated model to further improve the performance of federated model.
Good, Robust and Flexible System
Sophon P²C has good system robustness to response effectively with non-malicious errors and explicit attacks, further ensuring the security of the federated model. In addition, its system deployment is very flexible, and it supports entity deployment, container deployment, cloud deployment, edge deployment and other deployment methods, which can effectively reduce the human cost required for enterprise AI collaboration, and truly realize cost reduction and efficiency increase.
Application Scenarios / Cases
Credit Risk Management
Sophon P²C enhances credit risk management for the federated data network and helps consumer finance companies with their credit preliminary review. From the source of credit risk, it can help credit companies filter out credit blacklists or loan customers who have obviously not converted, and further reduce credit review costs in the later stages of the loan approval process.
Transwarp, Shaping the Future Data World
