Federated Learning
FEDERATED LEARNING
Federated Learning is a machine learning approach that enables model training across multiple decentralized devices or servers while keeping the data localized. In traditional machine learning, data is collected and sent to a central server where a model is trained. However, in many cases, the data might be sensitive, private, or too large to be efficiently transferred to a central location.
FED AVG & FED PROX
1. Federated Averaging: This is one of the most widely used aggregation methods. Each client trains its local model using its data and then sends the model’s parameters (weights and biases) to the central server. The central server computes the average of these parameters and updates the global model accordingly. Federated averaging balances contributions from different clients and is simple to implement.
2. Federated SGD (Stochastic Gradient Descent): Similar to traditional SGD, clients update the global model by applying local updates on their data. The central server aggregates the models by simply averaging the gradients sent by the clients.
DIFFERENTIAL PRIVACY
Differential privacy is a privacy-preserving technique that ensures the protection of sensitive data while allowing for useful insights to be extracted from the data. It adds noise to the data before sharing it to ensure that individual data points cannot be re-identified.
Federated learning, on the other hand, is a decentralized machine learning approach where multiple devices or servers collaboratively train a global model while keeping the data localized and not transferring the raw data to a central server. This helps in preserving privacy and reducing the risk of data breaches.
FOOLS GOLDS
“Fool’s gold” in the context of federated learning refers to the situation where participating devices or clients in a federated learning system send misleading or erroneous updates to the central server, which can adversely affect the model’s performance. This term is used to highlight the potential risks and challenges associated with federated learning, a decentralized machine learning approach.
