Question

The Lottery Ticket Hypothesis attempts to explain these constructs’ efficiency despite their low VC dimension. Proving (10[1])that a sequence of these constructs converges to any arbitrary function is the goal of several Universal Approximation Theorems. (10[1])One type of these constructs passes a 2D “mask” over values to compute an activation map. (-5[1])The LSTM was created to alleviate a problem with these constructs where a value (10[1])“vanishes” due to repeatedly applying the chain rule. Images can be processed using the “convolutional” (10[2])type (10[1])of these (10[2])constructs. (10[2])The weights of these constructs are typically updated using backpropagation, which was popularized by 2024 Nobel Laureate Geoffrey Hinton. For 10 points, “deep learning” uses what biologically-inspired computational constructs? ■END■ (10[1])

ANSWER: artificial neural networks [or deep neural networks or ANNs or DNNs; accept specific types of neural networks such as convolutional neural networks or recurrent neural networks or CNNs or RNNs; accept “The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks”; prompt on classifiers or learners or machine learning models or large language models or LLMs; prompt on artificial neurons; reject “biological neural networks”]
<Other Science>
= Average correct buzz position

Back to tossups

Buzzes

PlayerTeamOpponentBuzz PositionValue
Linus LuuCambridge BManchester1510
Omer KeskinOxford ACambridge D3410
Cormac StephensonSouthampton ACambridge A50-5
Kevin FlanaganBristol AVanderbilt6410
Robert BallantyneBristol BOxford B7910
Faiz AhmedBirminghamLSE B7910
Aisling SkeetDurham ASouthampton A8010
Michael WuSouthampton BDurham A8210
Justin KeungImperial ACambridge C8210
Danny FisherWarwick BLSE A8310
Chris LevesleyWarwick ADurham B8310
Agnijo BanerjeeCambridge ASouthampton A11310