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…The default behavior with batch normalized classifier networks is to use a running average of the activation moments at test time. Previous works (Radford et al., 2016) have instead used batch statistics when sampling images. While this is …
Alec Radford
scientist · 2 mentions across 1 reading
In this course
Radford is a primary architect of modern generative image synthesis, most notably through foundational work on Deep Convolutional GANs (DCGANs) that stabilized training of generative adversarial networks and enabled large-scale image generation. In the course readings, his 2016 work appears as a methodological reference point for handling batch normalization during generative sampling, establishing practical techniques that became standard in contemporary generative AI systems. His contributions exemplify how engineering choices in neural network design directly shape what kinds of computational creativity become feasible at scale.
Mentioned in 1 reading
Appears alongside
People mentioned in the same passages — sorted by co-occurrence weight.