Lemma 3. Consider the ReLU activation The proof of Theorem 1 is given below. The inequality 3 uses strictly monotone property of p () . Code is available at this link. The neural networks are updated using Adam with learning rate initializes at 0.035 and All of them have no communication constraints. The training time is shown in Table 1.

We provide an extended version of the Experimental Setup from Section 5 below. Linear Model This domain involves learning a linear model when the underlying mapping between features and predictions is cubic. Concretely, the aim is to choose the top B =1 out of N = 50 resources using a linear model. The fact that the features can be seen as 1-dimensional allows us to visualize the learned models (as seen in Figure 4). There are 200 ( x, y) pairs in each of the training and validation sets, and 400 ( x, y) pairs in the test set.

Carlo estimator, focusing on the Character V AE for molecular generation setting described in 5.3.1 and Importance sampling-based estimator (IS-MI) described in 3, and the naive Monte Carlo (MC-MI) equivalent. (Figure 1). The training dataset is comprised of 60k images and the test dataset is comprised of 10k images. No data augmentation is used at train time nor at inference. We jointly train a variational autoencoder with an auxiliary network (the "Property network") predicting digit thickness based on latent representation (see Figure 1).

ELBO objective (3) presented in the main text. Firstly, the latent variables have very different meanings. Another important contribution of the paper is the generalization of deep CAMA to generic measurement data. We also performed experiments using different DNN network architectures. Figure 15 shows the performance against different shifts.

Unitary recurrent neural networks (URNNs) have been proposed as a method to overcome the vanishing and exploding gradient problem in modeling data with long-term dependencies. A basic question is how restrictive is the unitary constraint on the possible input-output mappings of such a network? This works shows that for any contractive RNN with ReLU activations, there is a URNN with at most twice the number of hidden states and the identical input-output mapping. Hence, with ReLU activations, URNNs are as expressive as general RNNs. In contrast, for certain smooth activations, it is shown that the input-output mapping of an RNN cannot be matched with a URNN, even with an arbitrary number of states. The theoretical results are supported by experiments on modeling of slowly-varying dynamical systems.

Infinite width limit has shed light on generalization and optimization aspects of deep learning by establishing connections between neural networks and kernel methods. Despite their importance, the utility of these kernel methods was limited in large-scale learning settings due to their (super-)quadratic runtime and memory complexities. Moreover, most prior works on neural kernels have focused on the ReLU activation, mainly due to its popularity but also due to the difficulty of computing such kernels for general activations. In this work, we overcome such difficulties by providing methods to work with general activations. First, we compile and expand the list of activation functions admitting exact dual activation expressions to compute neural kernels.

Homomorphic Encryption (HE) is one of the most promising security solutions to emerging Machine Learning as a Service (MLaaS). Several Leveled-HE (LHE)-enabled Convolutional Neural Networks (LHECNNs) are proposed to implement MLaaS to avoid the large bootstrapping overhead. However, prior LHECNNs have to pay significant computational overhead but achieve only low inference accuracy, due to their polynomial approximation activations and poolings. Stacking many polynomial approximation activation layers in a network greatly reduces the inference accuracy, since the polynomial approximation activation errors lead to a low distortion of the output distribution of the next batch normalization layer. So the polynomial approximation activations and poolings have become the obstacle to a fast and accurate LHECNN model.

How can local-search methods such as stochastic gradient descent (SGD) avoid bad local minima in training multi-layer neural networks? Why can they fit random labels even given non-convex and non-smooth architectures? Most existing theory only covers networks with one hidden layer, so can we go deeper? In this paper, we focus on recurrent neural networks (RNNs) which are multi-layer networks widely used in natural language processing. They are harder to analyze than feedforward neural networks, because the \emph{same} recurrent unit is repeatedly applied across the entire time horizon of length $L$, which is analogous to feedforward networks of depth $L$. We show when the number of neurons is sufficiently large, meaning polynomial in the training data size and in $L$, then SGD is capable of minimizing the regression loss in the linear convergence rate. This gives theoretical evidence of how RNNs can memorize data. More importantly, in this paper we build general toolkits to analyze multi-layer networks with ReLU activations. For instance, we prove why ReLU activations can prevent exponential gradient explosion or vanishing, and build a perturbation theory to analyze first-order approximation of multi-layer networks.

Prior works for learning networks with ReLU activations assume that the bias ($b$) is zero. In order to deal with the presence of the bias terms, our proposed algorithm consists of robustly decomposing multiple higher order tensors arising from the Hermite expansion of the function $f(x)$. Using these ideas we also establish identifiability of the network parameters under very mild assumptions.

Rectified linear unit (ReLU) activations can also be thought of as'gates', which, either pass or stop their pre-activation input when they are'on' (when the pre-activation input is positive) or'off' (when the pre-activation input is negative) respectively. A deep neural network (DNN) with ReLU activations has many gates, and the on/off status of each gate changes across input examples as well as network weights. For a given input example, only a subset of gates are'active', i.e., on, and the sub-network of weights connected to these active gates is responsible for producing the output. At randomised initialisation, the active sub-network corresponding to a given input example is random. During training, as the weights are learnt, the active sub-networks are also learnt, and could hold valuable information.