New interpretation of attention in Transformers and Graph Attention Networks.
problem Understanding and improving attention mechanisms in deep learning models.
method Decomposed attention into a kernel and a normalization term; generalized the kernel function and norm.
result Generalized attention leads to better performance on various tasks.
Efficiently accelerates attention calculation for Transformers with relative positional encoding.
problem Quadratic complexity of attention in long sequences.
method Kernelized attention with Fast Fourier Transform (FFT) for RPE.
result Achieves O(n log n) time complexity, mitigates training instability, and outperforms other models.
Linear attention in Transformers can be interpreted as dynamic VAR models.
problem Misalignment between Transformers and autoregressive forecasting objectives.
method Interpreting linear attention as VAR, rearranging MLP, attention, and flow.
result SAMoVAR improves performance, interpretability, and efficiency.
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.
Transformers with linear space and time complexity for accurate attention estimation.
problem Efficiently estimating attention in large-scale tasks without relying on priors.
method Performers use Fast Attention Via positive Orthogonal Random features (FAVOR+) for linear approximation of softmax attention.
result Performers achieve competitive results on various tasks, demonstrating the effectiveness of their attention-learning approach.
The p-Laplacian Transformer improves transformer models by assigning higher attention weights to tokens in close proximity.
problem The self-attention mechanism in transformers does not effectively distinguish attention weights between tokens in close and non-close proximity.
method Proposes a novel class of transformers, p-Laplacian Transformers, that use p-Laplacian regularization to assign higher attention weights to tokens in close proximity. result Empirically demonstrates that p-Laplacian Transformers outperform baseline transformers on various benchmark datasets.
We incorporate Tensor-Product Representations within the Transformer in order to better support the explicit representation of relation structure. Our Tensor-Product Transformer (TP-Transformer) sets a new state of the art on the recently-introduced Mathematics Dataset containing 56 categories of free-form math word-pr…
Improved Transformer performance by addressing 'explaining away' effect.
problem Transformer's self-attention mechanism can explain away important input features.
method Proposed a doubly-normalized attention scheme to avoid 'explaining away' effect.
result Improved performance on benchmarks with the new attention scheme.
Linformer reduces transformer complexity to linear, improving efficiency.
problem High cost of training and deploying large transformer models for long sequences.
method Approximates self-attention with low-rank matrix, proposing Linformer with O(n) complexity. result Linformer performs similarly to standard transformers but is more memory- and time-efficient.
Transformer-MGK replaces redundant heads with Gaussian key mixtures, improving efficiency and performance.
problem Redundant attention heads in transformers degrade performance and efficiency.
method Transformer-MGK replaces redundant heads with a mixture of Gaussian keys.
result Transformer-MGK accelerates training and inference, reduces parameters and FLOPs, and achieves comparable or better accuracy.
ZeroS improves Transformers by adding negative weights, matching or beating softmax attention.
problem Limited performance of linear attention methods, especially in long context sequences.
method Proposes Zero-Sum Linear Attention (ZeroS) that removes the zero-order term and reweights zero-sum softmax residuals.
result ZeroS matches or exceeds standard softmax attention across various benchmarks, theoretically expanding representable functions.
A brain-inspired spiking Transformer reduces energy consumption and enhances interpretability.
problem Energy inefficiency and lack of interpretability in Transformer models.
method Spiking STDP Transformer using spike-timing-dependent plasticity (STDP) for self-attention.
result Achieves 94.35% and 78.08% accuracy on CIFAR-10 and CIFAR-100 datasets respectively, with 88.47% energy reduction.
Multi-head attention outperforms single-head in in-context linear regression tasks.
problem Comparing performance of transformer with single-/multi-head attention in in-context learning.
method Theoretical analysis of performance of transformers with different attention mechanisms in linear regression tasks.
result Multi-head attention with a substantial embedding dimension outperforms single-head attention in in-context linear regression tasks.
Attention mechanisms have become ubiquitous in NLP. Recent architectures, notably the Transformer, learn powerful context-aware word representations through layered, multi-headed attention. The multiple heads learn diverse types of word relationships. However, with standard softmax attention, all attention heads are de…
Transformers improve with Fourier integral attentions.
problem Inefficiency of dot-product attention in capturing feature dependencies.
method Interpreted attention as kernel regression, proposed FourierFormer with generalized Fourier integral kernels.
result FourierFormer achieves better accuracy and reduces redundancy.
Paper proposes Sinkformers for Transformers with doubly stochastic attention.
problem Improving Transformer models' accuracy in vision and natural language processing.
method Using Sinkhorn's algorithm to make attention matrices doubly stochastic instead of SoftMax normalization.
result Sinkformers enhance model accuracy in vision and natural language processing tasks.
Clustered attention improves transformer efficiency for large sequences.
problem Quadratic complexity of transformer attention matrix for large sequences.
method Group queries into clusters, compute attention only for centroids, and use centroids to approximate key/query dot products.
result Linear complexity with respect to sequence length for a fixed number of clusters.
The paper investigates polynomial alternatives to softmax in transformer models.
problem The effectiveness of softmax attention in transformers is questioned.
method The authors explore polynomial activations as alternatives to softmax, focusing on their ability to regularize the attention matrix.
result Certain polynomials can serve as effective substitutes for softmax in transformer applications, achieving strong performance.
Transformer is a powerful architecture that achieves superior performance on various sequence learning tasks, including neural machine translation, language understanding, and sequence prediction. At the core of the Transformer is the attention mechanism, which concurrently processes all inputs in the streams. In this …
SE(3)-Transformers maintain equivariance for 3D data under rotations and translations.
problem Ensuring stable and predictable performance in 3D data under transformations.
method Introducing a self-attention module that is equivariant under continuous 3D roto-translations.
result The SE(3)-Transformer outperforms non-equivariant and non-attention models on real-world datasets.
New attention model removes softmax, improving sequence length bias.
problem Softmax attention's limitations and sequence length bias.
method Replaces softmax with normalization in self-attention.
result Normalization leads to more robust and bias-free attention.
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.
Sparse Transformers can approximate dense Transformers with only O(n) connections.
problem Can sparse Transformers approximate arbitrary sequence-to-sequence functions?
method Proposed sufficient conditions for universal approximation and proved that sparse Transformers with O(n) connections can approximate dense models.
result Sparse Transformers with O(n) connections can approximate the same function class as dense models with n^2 connections.
Transformers become faster by linearizing self-attention.
problem Quadratic complexity of transformers makes them slow for long sequences.
method Expressed self-attention as a linear dot-product and used matrix product associativity to reduce complexity.
result Linear transformers are up to 4000x faster on long sequences.
ETC improves Transformer models for long and structured inputs.
problem Scaling input length and encoding structured inputs in Transformers.
method Introduces global-local attention, relative position encodings, and CPC pre-training.
result Achieves state-of-the-art results on four natural language datasets.
The Transformer is a fully attention-based alternative to recurrent networks that has achieved state-of-the-art results across a range of NLP tasks. In this paper, we analyze the structure of attention in a Transformer language model, the GPT-2 small pretrained model. We visualize attention for individual instances and…
We introduce the 2-simplicial Transformer, an extension of the Transformer which includes a form of higher-dimensional attention generalising the dot-product attention, and uses this attention to update entity representations with tensor products of value vectors. We show that this architecture is a useful inductive …
Centroid Transformers reduce memory and computation by summarizing inputs into centroids.
problem Efficiently summarize inputs with reduced memory and computation.
method Generalizes self-attention to map N inputs to M centroids (M ≤ N), reducing complexity.
result Centroid Transformers reduce memory and computation while preserving key information.
Transformers learn causal structure through gradient descent on self-attention mechanisms.
problem Understanding how transformers learn causal structure during training.
method In-context learning task and simplified two-layer transformer model.
result Gradient descent on a simplified transformer learns to encode latent causal graphs.
New analysis shows how attention masks and LayerNorm prevent rank collapse in transformers.
problem Rank collapse in transformer models with increasing depth.
method General analysis of rank collapse under self-attention, considering attention masks and LayerNorm.
result Self-attention with LayerNorm can prevent rank collapse and maintain a rich set of equilibria.
Study proposes a statistical test for Vision Transformer's attention mechanisms.
problem ViT's attention mechanisms may focus on irrelevant regions, leading to unreliable evidence.
method Selective inference framework to quantify statistical significance of attentions as p-values.
result Proposed method enables reliable quantification of false positive detection probability of attentions.
Transformer learns graph structure better with subgraph info.
problem Transformer struggles with structural similarity in graph learning.
method Structure-Aware Transformer with subgraph attention.
result Improves graph prediction benchmarks significantly.
Transformers cluster meaningless words around leaders for sentiment analysis.
problem Capturing context in sentiment analysis using transformers.
method Characterized transformers with hardmax self-attention and normalization, showing asymptotic convergence to clustered equilibrium.
result Transformers can effectively capture context by clustering meaningless words around leader words.
GTA improves transformer-based NVS models by encoding geometric structure.
problem Suboptimal positional encoding for 3D vision tasks.
method Geometry-aware attention mechanism encoding geometric structure of tokens.
result GTA improves learning efficiency and performance of NVS models.
Transformers interpreted as probabilistic Laplacian Eigenmaps steps.
problem Improving transformer performance through probabilistic interpretation.
method Probabilistic Laplacian Eigenmaps model derivation and graph diffusion step.
result Subtracting identity from attention matrix improves transformer performance.
Paper proposes a new hierarchical attention mechanism for multi-scale data.
problem Challenges in applying neural attention to multi-scale, multi-modal data.
method Developed a mathematical framework for multi-modal, multi-scale data and derived optimal neural attention mechanics.
result Proposed hierarchical attention mechanism improves transformer performance in multi-scale, multi-modal settings.
New method prevents entropy collapse in Transformer training, leading to more stable and robust models.
problem Training instability in Transformers, especially in attention layers.
method Spectral normalization with a learned scalar to prevent entropy collapse.
result Prevents entropy collapse, leading to more stable training.
Regularizes attention scores in vision transformers using bootstrapping.
problem Noisy and diffused attention maps in ViT limit interpretability.
method Statistical learning techniques, bootstrapping of attention scores.
result Improves shrinkage and sparsity of attention scores.
Transformers show strengths and weaknesses in complexity analysis.
problem Understanding the strengths and limitations of attention layers in transformers.
method Analysis of representation power through complexity parameters and task-specific constructions.
result Transformers can solve sparse averaging tasks with logarithmic complexity, but triple detection tasks require linear complexity.
This research integrates attention into XAI frameworks for better model explanations.
problem Improving the interpretability of transformer models.
method Developed two novel explanation methods: Shapley value decomposition and token-level directional derivatives.
result Attention weights can be meaningfully incorporated into XAI frameworks, enhancing transformer explainability.
Transformer models align words through attention weights, closely approximating Optimal Transport.
problem Understanding the internal mechanism of transformer models in language processing.
method Empirical evidence and theoretical analysis of attention weights and their relation to Optimal Transport.
result Transformer models can simulate gradient descent on the dual of entropy-regularized OT problem, providing a theoretical foundation for token alignment.
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.
Elliptical Attention improves transformer performance by focusing on contextually relevant features.
problem Transformer models suffer from representation collapse and are vulnerable to contaminated samples.
method Uses Mahalanobis distance to define hyper-ellipsoidal neighborhoods for attention weights.
result Elliptical Attention reduces representation collapse and enhances model robustness.
B-cos transformers explain Vision Transformers' decisions.
problem Lack of holistic explanations for transformer outputs.
method Formulate each component as dynamic linear, allowing a single linear transform for summarization.
result Bcos-ViTs are highly interpretable and competitive on ImageNet.
Study reveals Transformer's expressive power and mechanisms.
problem Understanding the approximation properties of Transformer for sequence modeling.
method Systematic study of Transformer's components and their combined effects, establishing approximation rates.
result Reveals roles of critical parameters in Transformer, such as number of layers and attention heads.
Paper proposes integrating wavelet transform, channel attention, and LSTM for better stock price prediction.
problem Inherently difficult stock price prediction due to low signal-to-noise ratio.
method Wavelet transform convolution, channel attention, and LSTM integration.
result Robust performance in post-pandemic market conditions.
Transformers learn to perform logistic regression in-context.
problem Understanding how transformers learn to perform specific tasks in-context.
method Constructed multi-layer transformers that perform in-context logistic regression through normalized gradient descent.
result Transformers can be trained to perform in-context logistic regression effectively.
Transformers mimic Bayesian reasoning in controlled settings, revealing geometric mechanisms.
problem Verifying if transformers perform Bayesian reasoning rigorously in natural data.
method Constructing Bayesian wind tunnels with known posteriors and proving memorization impossibility.
result Transformers achieve 10−3-10−4 bit accuracy in Bayesian posteriors, while MLPs fail.