The modern mathematics of deep learning

We describe the new field of mathematical analysis of deep learning. This field emerged around a list of research questions that were not answered within the classical framework of learning theory. These questions concern: the outstanding generalization power of overparametrized neural networks, the role of depth in deep architectures, the apparent absence of the curse of dimensionality, the surprisingly successful optimization performance despite the non-convexity of the problem, understanding what features are learned, why deep architectures perform exceptionally well in physical problems, and which fine aspects of an architecture affect the behavior of a learning task in which way.

Evaluate NER parsers with spaCy and Label Studio

This tutorial helps you evaluate accuracy of Named Entity Recognition (NER) taggers using Label Studio. Gather predictions from standard spaCY language models for a dataset based on transcripts from the podcast This American Life, then use Label Studio to correct the transcripts and determine which model performed better to focus future retraining efforts.

Good data Scientist, bad data scientist

At their core, data scientists exist to create business value with data. The ways in which value is generated will vary largely company to company, and even team to team. For teams that are early in their lifecycle, this will often take the form of building up basic analytical foundations. The team is likely blind to aspects of the product performance, user behaviour or total addressable market.

Diffusion models beat GANs on image synthesis

We show that diffusion models can achieve image sample quality superior to the current state-of-the-art generative models. We achieve this on unconditional image synthesis by finding a better architecture through a series of ablations.

Enhancing photorealism enhancement

We present an approach to enhancing the realism of synthetic images. The images are enhanced by a convolutional network that leverages intermediate representations produced by conventional rendering pipelines. The network is trained via a novel adversarial objective, which provides strong supervision at multiple perceptual levels. We analyze scene layout distributions in commonly used datasets and find that they differ in important ways. We hypothesize that this is one of the causes of strong artifacts that can be observed in the results of many prior methods.