A Generative Adversarial Network (GAN) involves Generator (G) and Discriminator (D) networks whose purpose, respectively, is to map random noise to samples and discriminate real and generated samples. Formally, the GAN objective, in its ori…One possible explanation is that this behavior is a consequence of D memorizing the training data, as suggested by experiments in Section 4.2. As it approaches perfect memorization, it receives less and less signal from real data, as both t…
Martin Arjovsky
scientist · 3 mentions across 2 readings
In this course
Arjovsky appears as a co-author of the foundational GAN framework (Goodfellow et al., 2014) that established the adversarial training paradigm central to generative modeling. His work enables the course's engagement with how competing neural networks—generator and discriminator—create a dynamic system for synthesizing data, a key mechanism in contemporary AI systems that the readings analyze through problems like discriminator memorization and training instability. Though cited primarily as part of the original GAN formulation, his contribution anchors discussions of how adversarial objectives structure machine learning systems and their pathological behaviors.
Mentioned in 2 readings
Appears alongside
People mentioned in the same passages — sorted by co-occurrence weight.
Adam Roberts 1Andrew Brock 1Ariel Herbert-Voss 1Dawn Song 1Eirikur Agustsson 1Eric Wallace 1Florian Tramer 1Holger Caesar 1Jasper R. R. Uijlings 1Jeff Donahue 1Karen Simonyan 1Katherine Lee 1Léon Bottou 1Matthew Jagielski 1Nicholas Carlini 1Radu Timofte 1Soumith Chintala 1Tom Brown 1Ulfar Erlingsson 1Vittorio Ferrari 1Alec Radford 1Casper Kaae Sønderby 1Ian Goodfellow 1Ishaan Gulrajani 1Junnan Lim 1Lars Mescheder 1Marc G. Bellemare 1Naveen Kodali 1Sebastian Nowozin 1Takeru Miyato 1