Casper Kaae Sønderby

TL;DR: A deep neural network-based approach that improves SP prediction across all domains of life and distinguishes between three types of prokaryotic SPs is presented.

TL;DR: This work presents a prediction algorithm using deep neural networks to predict protein subcellular localization relying only on sequence information, outperforming current state‐of‐the‐art algorithms, including those relying on homology information.

TL;DR: This article proposed a new inference model, the Ladder Variational Autoencoder, that recursively corrects the generative distribution by a data dependent approximate likelihood in a process similar to the recently proposed Ladder Network.

TL;DR: The accuracy of NetSurfP‐2.0 is assessed and it is found to consistently produce state‐of‐the‐art predictions for each of its output features, and the processing time has been optimized to allow predicting more than 1000 proteins in less than 2 hours, and complete proteomes in more than 1 day.

Abstract: Image super-resolution (SR) is an underdetermined inverse problem, where a large number of plausible high-resolution images can explain the same downsampled image. Most current single image SR methods use empirical risk minimisation, often with a pixel-wise mean squared error (MSE) loss. However, the outputs from such methods tend to be blurry, over-smoothed and generally appear implausible. A more desirable approach would employ Maximum a Posteriori (MAP) inference, preferring solutions that always have a high probability under the image prior, and thus appear more plausible. Direct MAP estimation for SR is non-trivial, as it requires us to build a model for the image prior from samples. Furthermore, MAP inference is often performed via optimisation-based iterative algorithms which don't compare well with the efficiency of neural-network-based alternatives. Here we introduce new methods for amortised MAP inference whereby we calculate the MAP estimate directly using a convolutional neural network. We first introduce a novel neural network architecture that performs a projection to the affine subspace of valid SR solutions ensuring that the high resolution output of the network is always consistent with the low resolution input. We show that, using this architecture, the amortised MAP inference problem reduces to minimising the cross-entropy between two distributions, similar to training generative models. We propose three methods to solve this optimisation problem: (1) Generative Adversarial Networks (GAN) (2) denoiser-guided SR which backpropagates gradient-estimates from denoising to train the network, and (3) a baseline method using a maximum-likelihood-trained image prior. Our experiments show that the GAN based approach performs best on real image data. Lastly, we establish a connection between GANs and amortised variational inference as in e.g. variational autoencoders.