Community Jameel

While neural networks used in practice are often very deep, the benefit of depth is not well understood. Interestingly, it is known that increasing depth is often harmful for regression tasks. In this work, we show that, in contrast to regression, very deep networks can be Bayes optimal for classification. In particular, this research provides simple and explicit activation functions that can be used with standard neural network architectures to achieve consistency. This work provides fundamental understanding of classification using deep neural networks, and the research team envisions it will help guide the design of future neural network architectures.

Wide and deep neural networks achieve consistency for classification

Details

author(s)

publication date

source

related programme

Link to publication

Generative AI in the era of 'alternative facts'

External data and AI are making each other more valuable

Removing biases from molecular representations via information maximisation

Effective human-AI teams via learned natural language rules and onboarding

A deep dive into single-cell RNA sequencing foundation models

Antibiotic identified by AI

LLM-grounded video diffusion models

Successful Development of a Natural Language Processing Algorithm for Pancreatic Neoplasms and Associated Histologic Features

Leveraging artificial intelligence in the fight against infectious diseases

BioAutoMATED: An end-to-end automated machine learning tool for explanation and design of biological sequences

Conformal language modeling

Comparison of mammography AI algorithms with a clinical risk model for 5-year breast cancer risk prediction: An observational study

Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii

Algorithmic pluralism: A structural approach towards equal opportunity

Artificial intelligence and machine learning in lung cancer screening

Wide and deep neural networks achieve consistency for classification

Autocatalytic base editing for RNA-responsive translational control

DiffDock: Diffusion steps, twists and turns for molecular docking

Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography

Sequential multi-dimensional self-supervised learning for clinical time series

Queueing theory: Classical and modern methods

Toward robust mammography-based models for breast cancer risk

The age of AI: And our human future

Uniform priors for data-efficient transfer

Machine learning under a modern optimisation lens

The marginal value of adaptive gradient methods in machine learning

Efficient graph-based image segmentation