Efficient Elastic Net Regularization with Sparsity - Zachary Chase Lipton

Yandex School of Data Analysis Conference Machine Learning: Prospects and Applications Sparse models are easily and quickly trained absent regularization. How- ever, working with large sparse datasets simultaneously requires regu- larization and efficiency. Elastic net, which generalizes L1 and squared L2 regularization, is a popular choice of regularizer because it simultaneously confers model sparsity similar to L1 regularizers and the generally superior performan