Subscribe to the gold package and get unlimited access to Shamra Academy
Register a new userExistence, uniqueness, and convergence rates for gradient flows in the training of artificial neural networks with ReLU activation
59
0
0.0
(
0
)
Added by
Adrian Riekert
Publication date
2021
fields
Informatics Engineering
and research's language is
English
Ask ChatGPT about the research
No Arabic abstract
The training of artificial neural networks (ANNs) with rectified linear unit (ReLU) activation via gradient descent (GD) type optimization schemes is nowadays a common industrially relevant procedure. Till this day in the scientific literature there is in general no mathematical convergence analysis which explains the numerical success of GD type optimization schemes in the training of ANNs with ReLU activation. GD type optimization schemes can be regarded as temporal discretization methods for the gradient flow (GF) differential equations associated to the considered optimization problem and, in view of this, it seems to be a natural direction of research to first aim to develop a mathematical convergence theory for time-continuous GF differential equations and, thereafter, to aim to extend such a time-continuous convergence theory to implementable time-discrete GD type optimization methods. In this article we establish two basic results for GF differential equations in the training of fully-connected feedforward ANNs with one hidden layer and ReLU activation. In the first main result of this article we establish in the training of such ANNs under the assumption that the probability distribution of the input data of the considered supervised learning problem is absolutely continuous with a bounded density function that every GF differential equation admits for every initial value a solution which is also unique among a suitable class of solutions. In the second main result of this article we prove in the training of such ANNs under the assumption that the target function and the density function of the probability distribution of the input data are piecewise polynomial that every non-divergent GF trajectory converges with an appropriate rate of convergence to a critical point and that the risk of the non-divergent GF trajectory converges with rate 1 to the risk of the critical point.
rate research
Read More
Deep neural networks (DNNs) have successfully learned useful data representations in various tasks, however, assessing the reliability of these representations remains a challenge. Deep Ensemble is widely considered the state-of-the-art method for uncertainty estimation, but it is very expensive to train and test. MC-Dropout is another alternative method, which is less expensive but lacks the diversity of predictions. To get more diverse predictions in less time, we introduce Randomized ReLU Activation (RRA) framework. Under the framework, we propose two strategies, MC-DropReLU and MC-RReLU, to estimate uncertainty. Instead of randomly dropping some neurons of the network as in MC-Dropout, the RRA framework adds randomness to the activation function module, making the outputs diverse. As far as we know, this is the first attempt to add randomness to the activation function module to generate predictive uncertainty. We analyze and compare the output diversity of MC-Dropout and our method from the variance perspective and obtain the relationship between the hyperparameters and output diversity in the two methods. Moreover, our method is simple to implement and does not need to modify the existing model. We experimentally validate the RRA framework on three widely used datasets, CIFAR10, CIFAR100, and TinyImageNet. The experiments demonstrate that our method has competitive performance but is more favorable in training time and memory requirements.
We study the expressive power of deep ReLU neural networks for approximating functions in dilated shift-invariant spaces, which are widely used in signal processing, image processing, communications and so on. Approximation error bounds are estimated with respect to the width and depth of neural networks. The network construction is based on the bit extraction and data-fitting capacity of deep neural networks. As applications of our main results, the approximation rates of classical function spaces such as Sobolev spaces and Besov spaces are obtained. We also give lower bounds of the $L^p (1le p le infty)$ approximation error for Sobolev spaces, which show that our construction of neural network is asymptotically optimal up to a logarithmic factor.
We explore convergence of deep neural networks with the popular ReLU activation function, as the depth of the networks tends to infinity. To this end, we introduce the notion of activation domains and activation matrices of a ReLU network. By replacing applications of the ReLU activation function by multiplications with activation matrices on activation domains, we obtain an explicit expression of the ReLU network. We then identify the convergence of the ReLU networks as convergence of a class of infinite products of matrices. Sufficient and necessary conditions for convergence of these infinite products of matrices are studied. As a result, we establish necessary conditions for ReLU networks to converge that the sequence of weight matrices converges to the identity matrix and the sequence of the bias vectors converges to zero as the depth of ReLU networks increases to infinity. Moreover, we obtain sufficient conditions in terms of the weight matrices and bias vectors at hidden layers for pointwise convergence of deep ReLU networks. These results provide mathematical insights to the design strategy of the well-known deep residual networks in image classification.
By employing a system of interacting stochastic particles as an approximation of the McKean--Vlasov equation and utilizing classical stochastic analysis tools, namely It^os formula and Kolmogorov--Chentsov continuity theorem, we prove the existence and uniqueness of strong solutions for a broad class of McKean--Vlasov equations. Considering an increasing number of particles in the approximating stochastic particle system, we also prove the $L^p$ strong convergence rate and derive the weak convergence rates using the Kolmogorov backward equation and variations of the stochastic particle system. Our convergence rates were verified by numerical experiments which also indicate that the assumptions made here and in the literature can be relaxed.
In this paper, we construct neural networks with ReLU, sine and $2^x$ as activation functions. For general continuous $f$ defined on $[0,1]^d$ with continuity modulus $omega_f(cdot)$, we construct ReLU-sine-$2^x$ networks that enjoy an approximation rate $mathcal{O}(omega_f(sqrt{d})cdot2^{-M}+omega_{f}left(frac{sqrt{d}}{N}right))$, where $M,Nin mathbb{N}^{+}$ denote the hyperparameters related to widths of the networks. As a consequence, we can construct ReLU-sine-$2^x$ network with the depth $5$ and width $maxleft{leftlceil2d^{3/2}left(frac{3mu}{epsilon}right)^{1/{alpha}}rightrceil,2leftlceillog_2frac{3mu d^{alpha/2}}{2epsilon}rightrceil+2right}$ that approximates $fin mathcal{H}_{mu}^{alpha}([0,1]^d)$ within a given tolerance $epsilon >0$ measured in $L^p$ norm $pin[1,infty)$, where $mathcal{H}_{mu}^{alpha}([0,1]^d)$ denotes the Holder continuous function class defined on $[0,1]^d$ with order $alpha in (0,1]$ and constant $mu > 0$. Therefore, the ReLU-sine-$2^x$ networks overcome the curse of dimensionality on $mathcal{H}_{mu}^{alpha}([0,1]^d)$. In addition to its supper expressive power, functions implemented by ReLU-sine-$2^x$ networks are (generalized) differentiable, enabling us to apply SGD to train.
suggested questions
Log in to be able to interact and post comments
comments
Fetching comments
Sign in to be able to follow your search criteria
