We propose an optimistic estimate to evaluate the best possible fitting
...
Dropout is a widely utilized regularization technique in the training of...
In this work, we study the mechanism underlying loss spikes observed dur...
Previous research has shown that fully-connected networks with small
ini...
The use of Physics-informed neural networks (PINNs) has shown promise in...
The phenomenon of distinct behaviors exhibited by neural networks under
...
Fractional diffusion equations have been an effective tool for modeling
...
Models with nonlinear architectures/parameterizations such as deep neura...
It is important to understand how the popular regularization method drop...
Unraveling the general structure underlying the loss landscapes of deep
...
Gradient descent or its variants are popular in training neural networks...
Substantial work indicates that the dynamics of neural networks (NNs) is...
In recent years, understanding the implicit regularization of neural net...
Understanding deep learning is increasingly emergent as it penetrates mo...
Machine learning has long been considered as a black box for predicting
...
A deep learning-based model reduction (DeePMR) method for simplifying
ch...
While deep learning algorithms demonstrate a great potential in scientif...
We prove a general Embedding Principle of loss landscape of deep neural
...
Although dropout has achieved great success in deep learning, little is ...
Complex design problems are common in the scientific and industrial fiel...
Empirical works suggest that various semantics emerge in the latent spac...
In this paper, we propose a model-operator-data network (MOD-Net) for so...
Understanding the structure of loss landscape of deep neural networks
(D...
It is important to study what implicit regularization is imposed on the ...
Deep neural network (DNN) usually learns the target function from low to...
Why heavily parameterized neural networks (NNs) do not overfit the data ...
Frequency perspective recently makes progress in understanding deep lear...
A supervised learning problem is to find a function in a hypothesis func...
Recent works show an intriguing phenomenon of Frequency Principle
(F-Pri...
It has been an important approach of using matrix completion to perform ...
Understanding the effect of depth in deep learning is a critical problem...
In this paper, we propose novel multi-scale DNNs (MscaleDNN) using the i...
How neural network behaves during the training over different choices of...
This paper aims at studying the difference between Ritz-Galerkin (R-G) m...
We focus on estimating a priori generalization error of two-layer ReLU
n...
In this paper, we propose the idea of radial scaling in frequency domain...
Along with fruitful applications of Deep Neural Networks (DNNs) to reali...
It remains a puzzle that why deep neural networks (DNNs), with more
para...
How different initializations and loss functions affect the learning of ...
We study the training process of Deep Neural Networks (DNNs) from the Fo...
Previous studies have shown that deep neural networks (DNNs) with common...
To understand how neural networks process information, it is important t...
In many realistic systems, maximum entropy principle (MEP) analysis prov...