Statistical Theory of Extreme Values and Some Practical Applications

By B. Brookes. 1955

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistics and Probability

Citations in the corpus

Predictive Sampling with Forecasting Autoregressive Models (2020, methodology): “One method to reparametrize categorical distributions is the Gumbel-Max trick (Gumbel, 1954), which has recently become popular in machine learning (Maddison et al., 2014; 2017; Jang et al., 2017; Kool et al., 2019).”
Categorical Reparameterization with Gumbel-Softmax (2016, methodology): “The Gumbel-Max trick (Gumbel, 1954; Maddison et al., 2014) provides a simple and efficient way to draw samples z from a categorical distribution with class probabilities π: z = one_hot ( arg max i [gi + log πi] ) (1) where g1...gk are i.i.d samples drawn from Gumbel(0, 1)1.”
Dynamic Neural Networks: A Survey (2021, methodology): “For instance, the gating functions in [46], [82] are both trained with the Gumbel SoftMax technique [224], [225] to control”
Autoregressive Image Generation without Vector Quantization (2024, methodology): “| Sampling from a categorical distribution can be approached by the Gumbel-max method [18] or inverse transform sampling.”
Meta-Recognition: The Theory and Practice of Recognition Score Analysis (2011, background): “Gumbel [24] showed that for any continuous and invertible initial distribution, only three models are needed, depending on whether the maximum or the minimum is of interest, and also if the observations are bounded from above or below.”
Bridging the Gap between Training and Inference for Neural Machine Translation (2019, methodology): “In practice, we can acquire more robust word-level oracles by introducing the Gumbel-Max technique (Gumbel, 1954; Maddison et al., 2014), which provides a simple and efficient way to sample from a categorical distribution.”
Dynamic Channel and Layer Gating in Convolutional Neural Networks (2020, methodology): “All the gates are trained using the Gumbel max trick[7] with sigmoid relaxation.”