Timezone: »

MLP-Mixer: An all-MLP Architecture for Vision
Ilya Tolstikhin · Neil Houlsby · Alexander Kolesnikov · Lucas Beyer · Xiaohua Zhai · Thomas Unterthiner · Jessica Yung · Andreas P Steiner · Daniel Keysers · Jakob Uszkoreit · Mario Lucic · Alexey Dosovitskiy

Wed Dec 08 12:30 AM -- 02:00 AM (PST) @ None #None

Convolutional Neural Networks (CNNs) are the go-to model for computer vision. Recently, attention-based networks, such as the Vision Transformer, have also become popular. In this paper we show that while convolutions and attention are both sufficient for good performance, neither of them are necessary. We present MLP-Mixer, an architecture based exclusively on multi-layer perceptrons (MLPs). MLP-Mixer contains two types of layers: one with MLPs applied independently to image patches (i.e. "mixing" the per-location features), and one with MLPs applied across patches (i.e. "mixing" spatial information). When trained on large datasets, or with modern regularization schemes, MLP-Mixer attains competitive scores on image classification benchmarks, with pre-training and inference cost comparable to state-of-the-art models. We hope that these results spark further research beyond the realms of well established CNNs and Transformers.

Author Information

Ilya Tolstikhin (Google, Brain Team, Zurich)
Neil Houlsby (Google)
Alexander Kolesnikov (Google)
Lucas Beyer (Google Brain Zürich)
Xiaohua Zhai (Google Brain)
Tom Unterthiner (Google Research)
Jessica Yung (Google)
ANDREAS P Steiner (Google)

- Education : medical doctor, MSc in bio-electronics - MD work computer-assisted diagnostics for tuberculosis screening in Tanzanian prisons - Developed next generation sequencing variant calling software and tools for epidemiological studies using hand-held devices - Software Engineer with Google since 2015, working on Shopping, ML, and ML productionization

Daniel Keysers (Google Research, Brain Team)
Jakob Uszkoreit (Google)
Mario Lucic (Google Brain)
Alexey Dosovitskiy (Inceptive)

More from the Same Authors