SGD

SGD most commonly refers to stochastic gradient descent, an iterative optimization algorithm used to minimize differentiable objective functions, particularly in machine learning. In SGD, model parameters are updated in the opposite direction of the gradient of the loss with respect to the parameters, computed on a subset of the training data rather than the full dataset. This typically involves a learning rate and a gradient estimate. Updates take the form theta := theta - eta * g. There are online versions (one example per update) and mini-batch versions (updates after processing a small batch of samples). SGD is scalable to large datasets and suitable for online learning; it introduces stochastic noise, which can help escape shallow local minima and improve generalization in some settings but can make convergence noisier. In practice, SGD is often combined with momentum or adaptive learning-rate schemes such as Adam or RMSprop, which modify the effective learning rate or incorporate past gradients. Proper choice of learning rate schedule, initialization, and regularization are important for performance. Applications include training deep neural networks, logistic regression, and other differentiable models.

SGD also stands for the Singapore dollar, the official currency of Singapore. The ISO 4217 code is SGD, and the symbol commonly used is S$. The currency is issued by the Monetary Authority of Singapore and is subdivided into 100 cents. It is widely used domestically and in international trade, and exchange-rate movements are influenced by Singapore's monetary policy framework, which emphasizes the exchange rate as the primary instrument of policy.