Self-attention models benefit equally from width and depth, but beyond a certain point, depth becomes less efficient.
problem Understanding the optimal balance between depth and width in self-attention models.
method Theoretical predictions and empirical ablations on networks of varying depths and widths.
result An optimal width of 30K is recommended for a 1-Trillion parameter network, marking a significant width for self-attention models.
Paper investigates Lipschitz constants of self-attention modules in neural networks.
problem Lipschitz constants of self-attention modules in neural networks.
method Proved standard dot-product self-attention is not Lipschitz for unbounded input domain. Proposed L2 self-attention that is Lipschitz. Derived upper bound on L2 self-attention's Lipschitz constant.
result Proved standard self-attention is not Lipschitz for unbounded input domain and proposed an alternative L2 self-attention that is Lipschitz.
GSA-Nets apply group equivariance to self-attention for vision tasks.
problem Improving self-attention networks for vision tasks.
method Define group-equivariant positional encodings.
result GSA-Nets outperform non-equivariant self-attention networks on vision benchmarks.
Self-attention prefers sparse functions of input sequences, reducing sample complexity.
problem Understanding the inductive biases of self-attention in modeling long-range dependencies.
method Theoretical analysis and synthetic experiments to probe sample complexity of learning sparse functions with Transformers.
result Bounded-norm Transformer networks can represent sparse functions of the input sequence with logarithmic sample complexity.
SANST uses self-attentive networks with spatial and temporal embeddings for better POI recommendations.
problem Next point-of-interest (POI) recommendation for users based on their history.
method SANST incorporates spatio-temporal patterns into self-attentive networks.
result SANST outperforms state-of-the-art models by up to 13.65% in nDCG@10.
TSAM predicts directed temporal links using GCN and self-attention.
problem Predicting links in directed temporal networks.
method GCN, self-attention mechanism, autoencoder architecture, graph attentional layers, graph convolutional layers, graph recurrent unit layer.
result TSAM outperforms benchmarks on four realistic networks.
Improved speech enhancement with MNTFA using time-frequency attention.
problem Speech enhancement with limited model size and memory.
method Designing MNTFA with self-attention modules for long sequences and joint training.
result MNTFA achieves better performance with fewer parameters than DPCRN.
A new neural network separates vocals from music accompaniment.
problem Separating vocals from music accompaniment in recordings.
method Self-attention convolutional neural network (CNN) with densely-connected blocks.
result 19.5% relative improvement in vocals separation.
Transformers' self-attention mechanism is mapped to a generalized Potts model.
problem Uncertainty in what type of data distribution self-attention can efficiently learn.
method Decouple word positions and embeddings, then show self-attention learns a generalized Potts model.
result Training self-attention is equivalent to solving the inverse Potts problem.
Analyzes self-attention in recurrent networks, proving it mitigates vanishing gradients.
problem Vanishing gradients in recurrent networks when capturing long-term dependencies.
method Formal analysis of self-attention's effect on gradient propagation, proposing a relevancy screening mechanism.
result Self-attention mitigates vanishing gradients in recurrent networks, providing guarantees.
The paper investigates how Global Self-attention improves GCNs.
problem Improving expressive power and addressing overfitting in GCNs.
method Applying Global Self-attention mechanism over graph features.
result GSA mechanism enhances expressive power and mitigates overfitting and over-smoothing in GCNs.
Derives an empirical capacity model for self-attention neural networks.
problem Theoretical capacity of large transformer models is not fully utilized by current optimization algorithms.
method Analyzes memory capacity of transformers using synthetic training data and common training algorithms.
result Derives an empirical capacity model (ECM) for a generic transformer.
LieTransformer extends self-attention to Lie groups for improved deep learning tasks.
problem Improving deep learning performance through group equivariant self-attention.
method LieSelfAttention layers that are equivariant to arbitrary Lie groups and their discrete subgroups.
result Competitive experimental results on various tasks.
Study reveals self-attention's role in learning and generalizing interactions.
problem Understanding self-attention's theoretical role in neural architectures.
method Interacting entities analysis, including multi-agent RL and genetic sequences.
result Self-attention efficiently represents, learns, and generalizes pairwise interactions.
Self-attention networks localize when eigenspectrum variance is small.
problem Self-attention mechanisms can lead to rank and entropy collapses, reducing model expressivity and trainability.
method Characterized attention localization using query-key eigenspectrum variance.
result Small eigenspectrum variance prevents both rank and entropy collapses, improving model performance.
Model detects electricity theft with high accuracy.
problem Detecting electricity theft on imbalanced datasets.
method Multi-head self-attention mechanism with dilated convolutions and binary mask.
result Achieved AUC of 0.926, improving previous work by 17%.
Learning latent representations of nodes in graphs is an important and ubiquitous task with widespread applications such as link prediction, node classification, and graph visualization. Previous methods on graph representation learning mainly focus on static graphs, however, many real-world graphs are dynamic and evol…
In this paper, we propose Dynamic Self-Attention (DSA), a new self-attention mechanism for sentence embedding. We design DSA by modifying dynamic routing in capsule network (Sabouretal.,2017) for natural language processing. DSA attends to informative words with a dynamic weight vector. We achieve new state-of-the-art …
Aligns attention distributions for improved accuracy and robustness.
problem Improving the accuracy and robustness of neural networks using attention mechanisms.
method Alignment attention that encourages key and query distributions to match within each head.
result Alignment attention leads to better accuracy, uncertainty estimation, and robustness across various tasks.
SANNE model generates embeddings for unseen nodes in graph networks.
problem Lack of embeddings for unseen nodes in graph networks.
method SANNE uses a transformer self-attention network to generate embeddings.
result SANNE achieves state-of-the-art results for node classification.
Quaternion self-attention reduces computational cost and improves performance.
problem Existing quaternion self-attention increases computational cost and diverges attention distributions.
method Proposes a shared-score quaternion self-attention mechanism.
result Reduces score-computation multiplications by 75% and softmax operations from four to one.
Insurance companies must manage millions of claims per year. While most of these claims are non-fraudulent, fraud detection is core for insurance companies. The ultimate goal is a predictive model to single out the fraudulent claims and pay out the non-fraudulent ones immediately. Modern machine learning methods are we…
Mathematical framework for understanding attention in neural networks.
problem Lack of theoretical understanding of attention in neural networks.
method Proposes a measure-theoretic model of attention and interprets self-attention as a system of self-interacting particles.
result Shows that attention is Lipschitz-continuous under suitable assumptions.
Transformer networks have lead to important progress in language modeling and machine translation. These models include two consecutive modules, a feed-forward layer and a self-attention layer. The latter allows the network to capture long term dependencies and are often regarded as the key ingredient in the success of…
Transformers predict price movements from limit order books.
problem Predicting price movements from limit order books.
method Causal convolutional network with masked self-attention.
result Significantly outperforms existing architectures on FI-2010 dataset.
Researchers improve transformer networks' optimization and understanding.
problem Improving the understanding and optimization of transformer networks.
method Introducing a convex alternative to the self-attention mechanism and reformulating the training problem as a convex optimization problem.
result Revealed an implicit regularization mechanism that promotes sparsity across tokens.
Self-attention model improves HAR from wearable sensors.
problem Capturing spatio-temporal context from sensor data.
method Proposes a self-attention based neural network model.
result Significant performance improvement over state-of-the-art models.
Modern deep learning approaches have achieved groundbreaking performance in modeling and classifying sequential data. Specifically, attention networks constitute the state-of-the-art paradigm for capturing long temporal dynamics. This paper examines the efficacy of this paradigm in the challenging task of emotion recog…
Sequential modelling with self-attention has achieved cutting edge performances in natural language processing. With advantages in model flexibility, computation complexity and interpretability, self-attention is gradually becoming a key component in event sequence models. However, like most other sequence models, self…
ST-SAN predicts flow with spatial-temporal dependencies using self-attention.
problem Challenges in predicting flow due to spatial-temporal dependencies.
method Spatial-Temporal Self-Attention Network (ST-SAN) that addresses temporal and spatial dependencies.
result Significant improvement in flow prediction accuracy (9% in inflow, 4% in outflow) compared to state-of-the-art methods.
The amount of available Earth observation data has increased dramatically in the recent years. Efficiently making use of the entire body information is a current challenge in remote sensing and demands for light-weight problem-agnostic models that do not require region- or problem-specific expert knowledge. End-to-end …
ConViT combines CNN and ViT strengths, improving image classification.
problem Combining the strengths of CNNs and ViTs while avoiding their limitations.
method Introducing GPSA, a form of positional self-attention with a soft convolutional inductive bias.
result ConViT outperforms DeiT on ImageNet while offering improved sample efficiency.
EFA extends self-attention to handle mixed data types and dynamic relevance.
problem Handling high-dimensional, mixed data types with dynamic relevance.
method Probabilistic generative model using self-attention and latent factor model.
result EFA consistently outperforms existing models in complex latent structure capture and reconstruction.
This paper offers an overview of neural network compression techniques.
problem Overparameterized neural networks are large and resource-intensive.
method Pruning, quantization, tensor decomposition, knowledge distillation.
result A comprehensive review of compression techniques for deep neural networks.
Improved electrical load forecasting model using Fourier-enhanced RNN.
problem Electrical load time series downscaling with high accuracy and low error.
method Combines recurrent neural network with Fourier seasonal embeddings and self-attention.
result Significantly reduces RMSE across different time horizons compared to existing methods.
Paper proposes multiscale self-attentive convolutions for vision and language.
problem Improving language and vision understanding models using self-attention.
method Developed 1D and 2D Self Attentive Convolutions (SAC), multiscale SAC (MSAC).
result MSAC enhances model performance for vision and language tasks.
Asynchronous events on the continuous time domain, e.g., social media actions and stock transactions, occur frequently in the world. The ability to recognize occurrence patterns of event sequences is crucial to predict which typeof events will happen next and when. A de facto standard mathematical framework to do this …
Active-memory mechanisms can replace self-attention in Transformers, but optimal results often require both.
problem Replacing self-attention with active-memory mechanisms in Transformers.
method Evaluation of various active-memory mechanisms in a Transformer model.
result Active-memory mechanisms can achieve comparable results to self-attention for language modeling, but optimal results are often achieved by combining both mechanisms.
A framework for transformer attention layers derived from SVR.
problem Developing principled attention mechanisms for transformers.
method Mapping self-attention to SVR, deriving new attention types.
result Improved transformer performance and efficiency.
Efficient neural network optimization reduces costs and improves model performance.
problem High computational costs in optimizing neural networks, especially at scale.
method Introduces self-attentive feed-forward neural units (SAFFU) for efficient optimization.
result Explicit solutions outperform models optimized by backpropagation alone, and further training with backpropagation leads to better optima from smaller data sets.
Study algebraic invariants from lightning self-attention models.
problem Understanding polynomial coefficients of self-attention mechanisms.
method Identify algebraic invariants using polynomial coefficients and coordinate geometry.
result Found linear and nonlinear families of algebraic invariants.
Random forests with attention and self-attention improve regression performance.
problem Improving regression model performance on various datasets.
method Proposes new models using attention and self-attention mechanisms to solve regression problems.
result The models improve model performance on many datasets.
Many real-world applications involve multivariate, geo-tagged time series data: at each location, multiple sensors record corresponding measurements. For example, air quality monitoring system records PM2.5, CO, etc. The resulting time-series data often has missing values due to device outages or communication errors. …
New model mimics neural next item recommendation using Hankel matrices.
problem Next item recommendation efficiency and structural knowledge capture.
method Tensor factorization with Hankel matrix representation.
result Model performs competitively with neural networks but is simpler.
Proposes a novel SAM operator for separate item and relational memories.
problem Limited memory interactions in neural networks.
method Introduces a Self-attentive Associative Memory (SAM) operator to separate item and relational memories.
result Achieves competitive results in various tasks, including geometry, graph, reinforcement learning, and question answering.
This work analyzes self-attention matrices using random matrix theory.
problem Understanding the theoretical behavior of self-attention layers in neural networks.
method Asymptotic spectral analysis of the attention matrix, Gaussian equivalence, and linearization.
result The singular value distribution of the attention matrix is asymptotically characterized by a linear model.
Kernel PCA explains self-attention mechanisms in deep learning models.
problem Understanding and explaining self-attention mechanisms in deep learning models.
method Deriving self-attention from kernel principal component analysis (kernel PCA).
result RPC-Attention, a robust attention mechanism, outperforms softmax attention in various tasks.
A method to generate long-range human actions by leveraging graph convolutional networks and self-attention.
problem Generating long-range skeleton-based human actions is challenging due to small frame deviations.
method Proposes a variant of GCNs with self-attention to adaptively sparsify action graphs and capture structure information.
result Extensive experiments show superior performance compared to existing methods on human action datasets.