This paper compares deeper and wider neural networks for optimal generalization error in Sobolev losses.
problem The dilemma of choosing between deeper or wider neural networks for optimal generalization error.
method Analytical investigations into the influence of sample points, parameters, and loss function regularity on neural network architecture.
result A higher number of parameters favors wider neural networks, while more sample points and greater loss function regularity favor deeper neural networks.
New approach to deeper graph neural networks to avoid performance degradation.
problem Performance degradation of graph neural networks when going deeper.
method Decoupling representation transformation and propagation in graph convolution operations.
result Deeper graph neural networks can be used to learn graph node representations from larger receptive fields.
Deeper neural networks learn lower frequency functions faster, according to a new principle.
problem Understanding why deeper learning is faster.
method Fourier analysis and filtering method to separate and analyze the frequency distribution of neural network outputs.
result Deeper hidden layers in neural networks bias towards lower frequency functions during training.
Deep neural networks achieve state-of-the-art performance for a range of classification and inference tasks. However, the use of stochastic gradient descent combined with the nonconvexity of the underlying optimization problems renders parameter learning susceptible to initialization. To address this issue, a variety o…
We establish, for the first time, connections between feedforward neural networks with ReLU activation and tropical geometry --- we show that the family of such neural networks is equivalent to the family of tropical rational maps. Among other things, we deduce that feedforward ReLU neural networks with one hidden laye…
In this paper, we study the problem of node representation learning with graph neural networks. We present a graph neural network class named recurrent graph neural network (RGNN), that address the shortcomings of prior methods. By using recurrent units to capture the long-term dependency across layers, our methods can…
It is well-known that neural networks are universal approximators, but that deeper networks tend in practice to be more powerful than shallower ones. We shed light on this by proving that the total number of neurons m required to approximate natural classes of multivariate polynomials of n variables grows only line…
Differentiable NAS frameworks grow networks wider and deeper, revealing biases in wiring evolution.
problem Understanding the evolution of neural architecture wiring in differentiable NAS methods.
method Unified view on searching algorithms, local cost minimization, empirical and theoretical analyses.
result Implicit inductive biases cause observed searching patterns in differentiable NAS methods.
This paper investigates how forgetting affects neural network representations and stabilizes deeper layers.
problem Catastrophic forgetting in machine learning models trained on sequential tasks.
method Representational analysis techniques and empirical studies on CIFAR-10 and CIFAR-100 datasets.
result Deeper layers are disproportionately the source of forgetting, and methods to mitigate forgetting stabilize these layers.
Deeper networks are better for local labels, but shallower for global labels.
problem Understanding the effect of depth in overparameterized neural networks.
method Introduced local and global labels to investigate the advantage of depth.
result Deeper networks are better for local labels, shallower for global labels.
Distributed implementations of mini-batch stochastic gradient descent (SGD) suffer from communication overheads, attributed to the high frequency of gradient updates inherent in small-batch training. Training with large batches can reduce these overheads; however, large batches can affect the convergence properties and…
Why does Deep Learning work? What representations does it capture? How do higher-order representations emerge? We study these questions from the perspective of group theory, thereby opening a new approach towards a theory of Deep learning. One factor behind the recent resurgence of the subject is a key algorithmic step…
Why does Deep Learning work? What representations does it capture? How do higher-order representations emerge? We study these questions from the perspective of group theory, thereby opening a new approach towards a theory of Deep learning. One factor behind the recent resurgence of the subject is a key algorithmic step…
A longstanding problem for Deep Neural Networks (DNNs) is understanding their puzzling ability to generalize well. We approach this problem through the unconventional angle of \textit{cognitive abstraction mechanisms}, drawing inspiration from recent neuroscience work, allowing us to define the Cognitive Neural Activat…
Langevin algorithms improve training of very deep neural networks, especially for image classification.
problem Training very deep neural networks is challenging due to increased non-linearity and the risk of getting stuck in local minima.
method Comparison of Langevin and non-Langevin algorithms for training deep neural networks, introduction of Layer Langevin algorithm.
result Langevin algorithms, especially Layer Langevin, lead to significant improvements in training deep neural networks, particularly for image classification tasks.
Deep neural networks are over-parameterized, which implies that the number of parameters are much larger than the number of samples used to train the network. Even in such a regime deep architectures do not overfit. This phenomenon is an active area of research and many theories have been proposed trying to understand …
Pruning is a well-established technique for removing unnecessary structure from neural networks after training to improve the performance of inference. Several recent results have explored the possibility of pruning at initialization time to provide similar benefits during training. In particular, the "lottery ticket h…
Analytical method finds deeper optima in two-layer ReLU networks.
problem Training two-layer ReLU networks with analytical methods.
method Analytically finding critical points of the loss function for one layer while keeping the other fixed.
result Significantly smaller training loss values on real datasets compared to gradient descent methods.
Deep learning is finding its way into the embedded world with applications such as autonomous driving, smart sensors and aug- mented reality. However, the computation of deep neural networks is demanding in energy, compute power and memory. Various approaches have been investigated to reduce the necessary resources, on…
In recent times, many of the breakthroughs in various vision-related tasks have revolved around improving learning of deep models; these methods have ranged from network architectural improvements such as Residual Networks, to various forms of regularisation such as Batch Normalisation. In essence, many of these techni…
Recently, deep residual networks have been successfully applied in many computer vision and natural language processing tasks, pushing the state-of-the-art performance with deeper and wider architectures. In this work, we interpret deep residual networks as ordinary differential equations (ODEs), which have long been s…
New findings show learning deeper neural networks is hard even with Gaussian inputs and non-degenerate weights.
problem The computational complexity of learning neural networks, especially deeper ones.
method Smoothed analysis framework and local pseudorandom generators.
result Learning depth-3 ReLU networks under Gaussian input distribution is hard even if weight matrices are non-degenerate.
Network degeneracy affects training performance, especially in deep networks.
problem Degeneracy in deep neural networks leads to poor training performance.
method Predicted degeneracy level correlates with training dynamics using finite and infinite width networks.
result Degeneracy in neural networks correlates with training performance and can be predicted.
New approach to understand recurrent policies as FSMs without minimization.
problem Minimization of FSMs obscures the semantics of policy decisions.
method Start with unminimized FSM, apply interpretable reductions, use attention tool.
result Reveals insights into policy decisions not previously noticed.
Deeper models have a more favorable optimization landscape, making them more robust to noise.
problem Characterizing the effect of depth on the optimization landscape of linear regression models.
method Robust and over-parameterized setting, simple sub-gradient method.
result A simple sub-gradient method converges to a balanced solution that is close to the ground truth and enjoys a flat local landscape.
Convolutional neural networks have been widely deployed in various application scenarios. In order to extend the applications' boundaries to some accuracy-crucial domains, researchers have been investigating approaches to boost accuracy through either deeper or wider network structures, which brings with them the expon…
This work proposes splitting deep neural networks into smaller sub-networks for faster and more efficient distillation.
problem Challenges in training deep neural networks, including local optima, gradient issues, and computational demands.
method Proposes a non-end-to-end distillation approach by splitting networks into smaller, independent sub-networks (neighbourhoods).
result Independent training of smaller sub-networks can speed up distillation and improve efficiency in various applications.
The paper explores stability and generalization of deep GCNs.
problem Understanding the stability and generalization of deep GCNs from a theoretical perspective.
method Theoretical analysis of stability and generalization properties of deep GCNs.
result The stability and generalization of deep GCNs are influenced by the maximum absolute eigenvalue of the graph filter operators and the depth of the network.
DeeperGCN tackles deep GCNs by overcoming vanishing gradient and over-smoothing issues.
problem Vanishing gradient, over-smoothing, and over-fitting issues in deep GCNs.
method DeeperGCN uses differentiable generalized aggregation functions and a novel normalization layer (MsgNorm) to train deep GCNs.
result DeeperGCN significantly boosts performance on large-scale graph learning tasks.
Proposes a new information-theoretic framework for analyzing deep neural networks.
problem Difficulty in analyzing deep neural networks using existing theoretical frameworks.
method Introduces an information-theoretic framework with new notions of regret and sample complexity.
result Establishes sample complexity bounds for deep neural networks that are width-independent and linear in depth.
Echo state networks are computationally lightweight reservoir models inspired by the random projections observed in cortical circuitry. As interest in reservoir computing has grown, networks have become deeper and more intricate. While these networks are increasingly applied to nontrivial forecasting tasks, there is a …
This paper relates parameter distance to gradient breakdown for a broad class of nonlinear compositional functions. The analysis leads to a new distance function called deep relative trust and a descent lemma for neural networks. Since the resulting learning rule seems to require little to no learning rate tuning, it m…
Study shows LLMs can remove half of layers without significant performance drop.
problem Understanding knowledge storage in LLMs' weights.
method Layer pruning and finetuning to identify and remove unnecessary parameters.
result Minimal degradation of performance after removing up to half of layers.
Giving provable guarantees for learning neural networks is a core challenge of machine learning theory. Most prior work gives parameter recovery guarantees for one hidden layer networks, however, the networks used in practice have multiple non-linear layers. In this work, we show how we can strengthen such results to d…
Study uses complex networks and machine learning to predict soccer match outcomes.
problem Predicting soccer match outcomes with complex networks and machine learning.
method Complex network metrics and match statistics were used to build machine learning models.
result Models based on passing networks were as effective as traditional models using match statistics.
The performance of graph neural nets (GNNs) is known to gradually decrease with increasing number of layers. This decay is partly attributed to oversmoothing, where repeated graph convolutions eventually make node embeddings indistinguishable. We take a closer look at two different interpretations, aiming to quantify o…
We provide several new depth-based separation results for feed-forward neural networks, proving that various types of simple and natural functions can be better approximated using deeper networks than shallower ones, even if the shallower networks are much larger. This includes indicators of balls and ellipses; non-lin…
A promising paradigm for achieving highly efficient deep neural networks is the idea of evolutionary deep intelligence, which mimics biological evolution processes to progressively synthesize more efficient networks. A crucial design factor in evolutionary deep intelligence is the genetic encoding scheme used to simula…
We develop a general duality between neural networks and compositional kernels, striving towards a better understanding of deep learning. We show that initial representations generated by common random initializations are sufficiently rich to express all functions in the dual kernel space. Hence, though the training ob…
Many interesting problems in machine learning are being revisited with new deep learning tools. For graph-based semisupervised learning, a recent important development is graph convolutional networks (GCNs), which nicely integrate local vertex features and graph topology in the convolutional layers. Although the GCN mo…
We propose a novel approach for sampling realistic financial correlation matrices. This approach is based on generative adversarial networks. Experiments demonstrate that generative adversarial networks are able to recover most of the known stylized facts about empirical correlation matrices estimated on asset returns.…
A new technique normalizes nodes within groups to improve GNN performance.
problem Over-smoothing in deeper GNNs reduces node distinguishability.
method Differentiable group normalization (DGN) to separate node distributions among groups.
result DGN makes GNN models more robust to over-smoothing and achieves better performance with deeper GNNs.
We develop a statistical mechanical approach based on the replica method to study the design space of deep and wide neural networks constrained to meet a large number of training data. Specifically, we analyze the configuration space of the synaptic weights and neurons in the hidden layers in a simple feed-forward perc…
Activation functions play a key role in neural networks so it becomes fundamental to understand their advantages and disadvantages in order to achieve better performances. This paper will first introduce common types of non linear activation functions that are alternative to the well known sigmoid function and then eva…
In this paper, we prove a conjecture published in 1989 and also partially address an open problem announced at the Conference on Learning Theory (COLT) 2015. With no unrealistic assumption, we first prove the following statements for the squared loss function of deep linear neural networks with any depth and any widths…
Nonlocal neural networks have been proposed and shown to be effective in several computer vision tasks, where the nonlocal operations can directly capture long-range dependencies in the feature space. In this paper, we study the nature of diffusion and damping effect of nonlocal networks by doing spectrum analysis on t…
Iterative learning to infer approaches have become popular solvers for inverse problems. However, their memory requirements during training grow linearly with model depth, limiting in practice model expressiveness. In this work, we propose an iterative inverse model with constant memory that relies on invertible networ…
Influence functions are inaccurate in deep learning models, especially for deeper networks.
problem Inaccuracies in influence functions in deep learning models.
method Empirical study of influence functions in neural network models trained on various datasets.
result Influence estimates are often erroneous for deeper networks and require regularization.