When we lowered prevalence at the test node, all performance parameters, including the F1 score (a measure of accuracy), were more resistant for SL than for individual nodes (Extended Data Fig. 7f–j). Finally, the nodes use the local data to evaluate the model with the updated parameter values and create evaluation criteria. When all merge participants are complete, the leader merges the local evaluation criteria values to calculate the number of global evaluation time criteria. Inspired by biology, swarm learning is based on blockchain, and is designed to ensure that only legitimate participants join a decentralized learning network. This allows companies to leverage distributed data while protecting data privacy and security.
We observe that MSL can also obtain better model accuracy in PAMAP experiment, even though the local model accuracy of each SLN is relatively low. The reason is that the neural network used in the experiment is more adept at processing image data, like the COVID-19 dataset, while the type of COVID-19 is sensor-based time series. We also observe that the accuracy of the global model obtained by MSL approximates the accuracy of the central learning model. Through the above evaluation, we can observe that MSL can improve the cognitive accuracy of local models, both for cognitive image data and cognitive time series data of healthcare. Although the experimental results show that the accuracy of MSL model decreases slightly as the number of SLNs involved in training increases.
2 Informal security analysis
At the time, HPE promised to combine Determined AI’s software solution with HPE’s own AI and HPC offerings—and now with the MLDS, it’s doing exactly that. For example, when able to share fraud-related learnings with numerous financial institutions at once, banking and financial services institutions can fight the expected global loss of more than $400 billion in credit card fraud over the next decade. It’s inefficient to move large volumes of data back and forth from the same source and the method is becoming less and less supportive of devices that are growing increasingly more intelligent by the day. Data privacy and ownership rules and regulations also limit data sharing and movement. Birds, fish, insects and other creatures have long been known to swarm or “murmur,” — which means on occasion they will move and coordinate as a group, rather than as directed by a centralized leader.
- Data is shuttled back and forth from the edge to a central data center for analysis.
- Data are kept locally and local confidentiality issues are addressed26, but model parameters are still handled by central custodians, which concentrates power.
- Swarm learning, developed jointly by HPE and DZNE (The German Center for Neurodegenerative Diseases), is opening up exciting new possibilities for collaboration, and greatly accelerating medical research.
- The model is configured for training with Adam optimization and to compute the binary cross-entropy loss between true labels and predicted labels.
- Indeed, statements by lawmakers have emphasized that privacy rules apply fully during a pandemic43.
Differences in performance metrics were tested using the one-sided Wilcoxon signed rank test with continuity correction. In the fourth use case, we addressed whether SL could be used to detect individuals with COVID-19 (Fig. 1k, Supplementary Table 6). Our team of HPE and other technology experts shares insights about relevant topics related to artificial intelligence, data analytics, IoT, and telco. The MLDS offers a full software and services stack, including a training platform (the HPE Machine Learning Development Environment), container management (Docker), cluster management (HPE Cluster Manager) and Red Hat Enterprise Linux. In a one-two punch of new HPC-backed AI announcements, Hewlett Packard Enterprise (HPE) today announced its new Machine Learning Development System (MLDS) and Swarm Learning solutions. Both are aimed at easing the burdens of AI development in a development environment that increasingly features large amounts of protected data and specialized hardware.
Supplementary Table 6
Detailed descriptions of the SLL, the architecture principles, the SL process, implementation, and the environment can be found in the Supplementary Information. First, we used peripheral blood mononuclear cell (PBMC) transcriptomes from more than 12,000 individuals (Fig. 1f–h) in three datasets (A1–A3, comprising two types of microarray and RNA sequencing (RNA-seq))3. If not otherwise stated, we used sequential deep neural networks with default settings28. For each real-world scenario, samples were split into non-overlapping training datasets and a global test dataset29 that was used for testing the models built at individual nodes and by SL (Fig. 2a). Within training data, samples were ‘siloed’ at each of the Swarm nodes in different distributions, thereby mimicking clinically relevant scenarios (Supplementary Table 1). As cases, we used samples from individuals with acute myeloid leukaemia (AML); all other samples were termed ‘controls’.
An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies. Our aim is to bring value to your business by delivering apprenticeships as well as training to upskill your current workforce so as to develop well-rounded, well-trained professionals ready for that next step in their career journey. We also offer a quick and efficient recruitment process with professional guidance every step of the way. In this section, we set a series of experiments to verify the accuracy of our system. Based on the above assumptions, MSL will be vulnerable to the following two attack models. Investigators were not blinded to allocation during experiments and outcome assessment.
Differentially private knowledge transfer for federated learning
A new node enrolls via a blockchain smart contract, obtains the model, and performs local model training until defined conditions for synchronization are met (Extended Data Fig. 1c). Next, model parameters are exchanged via a Swarm application programming interface (API) and merged to create an updated model with updated parameter settings before starting a new training round (Supplementary Information). Fast and reliable detection of patients with severe and heterogeneous illnesses is a major goal of precision medicine1,2. Patients with leukaemia can be identified using machine learning on the basis of their blood transcriptomes3. However, there is an increasing divide between what is technically possible and what is allowed, because of privacy legislation4,5.
The SLNs are practical privacy healthcare data owners and model training nodes with anonymous identities. There are semi-honest SLNs in MSL system, and they will perform the negotiated process for the global model but may try to infer the privacy of healthcare data from other honest SLNs. The other semi-honest SLNs are also excepted to gain a disproportionate benefit from the final healthcare global model. In addition, the communication channels between SLN and SNN are secure, and the information SLN obtains from the PBN is untampered. First, we distributed cases and controls unevenly at and between nodes (dataset A2) (Fig. 2b, Extended Data Fig. 2a, Supplementary Information), and found that SL outperformed each of the nodes (Fig. 2b).
This by overcoming challenges in data privacy and ownership and operational inefficiencies — that is, medical images are large and duplicating them is often simply out of the question. Having more data and more diverse data has resulted in greater accuracy in disease classification, Hotard said. The tool is designed to provide users with containers that are easily integrated with AI models via the HPE swarm API.
Then, we integrated DAO with the permissioned blockchain to incentivize institutions or individuals with high-quality data resources to participate in metaverse swarm learning and guarantee that all SLNs can fairly benefit from the process of model sharing. Security analysis showed that our proposed swarm learning-based model sharing could enable privacy protection, model quality control, swarm learning services and improve security and fairness without requiring centralized trust. Numerical results on two public healthcare datasets verified that our proposed model sharing could do collaborative training with privacy protection and obtain higher model accuracy than local training. SL builds on two proven technologies, distributed machine learning and blockchain (Supplementary Information).
Business Development Intern
A decentralized, privacy-preserving ML framework utilizes the computing power at, or near, the distributed data sources to run the ML algorithms that train the models. Training the model occurs at the edge where data is most recent, where accurate, and data-driven decisions are necessary. We built a second use case to identify patients with tuberculosis (TB) from blood transcriptomes30,31 (Fig. 1i, Supplementary Information). First, we used all TB samples (latent and active) as cases and distributed TB cases and controls evenly among the nodes (Extended Data Fig. 7a). SL outperformed individual nodes and performed slightly better than a central model under these conditions (Extended Data Fig. 7b, Supplementary Information).
For example, doctors can use the metaverse to discover relevant healthcare information from patient data through AI algorithms and assist in clinical decision-making . Besides decision-making, metaverse can stratify disease, reduce healthcare, and improve productivity in the virtual world . Metaverse applications in healthcare can realize patient access triage and further reduce the burden on the healthcare system, thereby directing scarce healthcare resources to patients with the most urgent healthcare requirements. Patient samples were provided by P.P., N.A.A., S.K., F.T., M. Bitzer, S.O., N.C., C.H., D.P., U.B., F.K., T.F., P.S, C.L., M.A., J.R., B.K., M.S., J.H., S.S., S.K.-H., J.N., D.S., I.K., A.K., R.B., M.G.N., M.M.B.B., E.J.G.-B, and M.K.
The Swarm Learning framework
My tutor Suzanne provided knowledge and support throughout, especially during the End-Point Assessment process. A big thank you to Suzanne for supporting me throughout the apprenticeship and getting me across the line. This file contains a more detailed description of Swarm Learning and the scenarios that were used for evaluation, as well as a Supplementary Discussion. Swarm learning minimizes the risk of regulatory compliance by eliminating the need to transfer real data, and solves boundary challenges by providing training close to the data. “Consider, for example, three breast cancer research institutions in the US, Europe, and Asia, each with its own limited dataset on breast cancer.
Swarm Learning to identify tuberculosis
The word “swarm” was inspired by the fact that various animals, usually for their own protection, exhibit a kind of decentralized behavior unrelated to the movements of the leader of their flock. HPE pitches the Machine Learning Development System as an end-to-end solution purpose-built for AI, stretching from software to hardware. The origins of the MLDS stretch back nearly a year to HPE’s acquisition of Determined AI, developer of a software stack for training AI models faster at scale.
HPE Swarm Learning demo
Due to the differences between the metaverse and the physical healthcare world, applying swarm learning to the metaverse healthcare faces two main challenges. One is that the nodes with swarm learning ability in the physical healthcare world are often endorsed by fundamental healthcare institutions, which would have more anonymous avatars in the metaverse. Anonymous identities in the model sharing process of swarm learning will bring security issues.