68 datasets found

The Hamiltonian neural network architecture is used to solve DE systems. The Hamiltonian NN is an evolution of previously used unsupervised NNs for finding solutions to DEs that...

The dataset is used to study the computational complexity of learning pushforwards of Gaussians under one-hidden-layer ReLU networks.

This work addresses the problem of planting and defending cryptographic-based backdoors in artificial intelligence (AI) models. The motivation comes from our lack of...

PROBEN1 is a set of neural network problems and benchmarking rules.

The dataset is used for analyzing individual neurons in pre-trained models.

The dataset used in the paper is related to neural networks and artificial intelligence.

Energy-based out-of-distribution detection.

Enhancing the reliability of out-of-distribution image detection in neural networks.

A baseline for detecting misclassified and out-of-distribution examples in neural networks.

The dataset consists of 24 state-of-the-art Google edge neural network models, including CNNs, LSTMs, Transducers, and RCNNs.

Zenkai is an open-source framework designed to give researchers more control and flexibility over building and training deep learning machines.

The dataset used in the paper is a set of embodied recurrent neural networks that perform object categorization and pole-balancing tasks.

The dataset used in the paper is a neural network with L layers numbered 1,..., L, in which each of the hidden layers has N neurons.

The dataset used in this paper is MNIST, CIFAR10, and ImageNet.

The ONNX dataset is used to test the onnx-mlir compiler.

The HEE model uses CIFAR10 to demonstrate its ability to generate high-quality images.

The HEE model uses 2D synthetic datasets and FashionMNIST to validate its capabilities.

The dataset is not explicitly described in the paper, but it is mentioned that the authors study the extent to which wide neural networks may be approximated by Gaussian processes.

The dataset used in this paper is a collection of training data for over-parameterized deep neural networks.

This paper develops an approach to learn a policy of a dynamical system that is guaranteed to be both provably safe and goal-reaching.