Efficient Representation of Natural Image Patches
This work addresses fundamental computational challenges in early visual systems and offers potential efficiency improvements for deep learning models, though it is incremental in its approach.
The paper tackles the problem of achieving efficient information transmission and accurate probability distribution modeling in early visual systems, proving that optimizing for one does not guarantee the other, and demonstrates that a nonlinear population code with biologically plausible loss functions can realize efficient representations, showing a significant efficiency advantage over a contemporary deep learning model.
Utilizing an abstract information processing model based on minimal yet realistic assumptions inspired by biological systems, we study how to achieve the early visual system's two ultimate objectives: efficient information transmission and accurate sensor probability distribution modeling. We prove that optimizing for information transmission does not guarantee optimal probability distribution modeling in general. We illustrate, using a two-pixel (2D) system and image patches, that an efficient representation can be realized through a nonlinear population code driven by two types of biologically plausible loss functions that depend solely on output. After unsupervised learning, our abstract information processing model bears remarkable resemblances to biological systems, despite not mimicking many features of real neurons, such as spiking activity. A preliminary comparison with a contemporary deep learning model suggests that our model offers a significant efficiency advantage. Our model provides novel insights into the computational theory of early visual systems as well as a potential new approach to enhance the efficiency of deep learning models.