How transformers learn structured data: insights from hierarchical filtering

• URL: http://arxiv.org/abs/2408.15138v2
• Date: Mon, 09 Dec 2024 16:53:42 GMT
• Title: How transformers learn structured data: insights from hierarchical filtering
• Authors: Jerome Garnier-Brun, Marc Mézard, Emanuele Moscato, Luca Saglietti,
• Abstract summary: We introduce a hierarchical filtering procedure for generative models of sequences on trees. We provide evidence that vanilla encoder-only transformers can approximate the exact inference algorithm when trained on root classification. We find clear evidence of a reconstruction of correlations on successive length scales corresponding to the various levels of the hierarchy.
• Score: 2.7784685368355744
• License:
• Abstract: Understanding the learning process and the embedded computation in transformers is becoming a central goal for the development of interpretable AI. In the present study, we introduce a hierarchical filtering procedure for generative models of sequences on trees, allowing us to hand-tune the range of positional correlations in the data. Leveraging this controlled setting, we provide evidence that vanilla encoder-only transformers can approximate the exact inference algorithm when trained on root classification and masked language modeling tasks, and study how this computation is discovered and implemented. We find that correlations at larger distances, corresponding to increasing layers of the hierarchy, are sequentially included by the network during training. Moreover, by comparing attention maps from models trained with varying degrees of filtering and by probing the different encoder levels, we find clear evidence of a reconstruction of correlations on successive length scales corresponding to the various levels of the hierarchy, which we relate to a plausible implementation of the exact inference algorithm within the same architecture.

Related papers

• Encoding Agent Trajectories as Representations with Sequence Transformers [0.4999814847776097]
  We propose a model for representing high dimensional trajectories with neural-based network architecture. Similar to language models, our Transformer Sequence for Agent temporal Representations (STARE) model can learn representations and structure in trajectory data. We present experimental results on various synthetic and real trajectory datasets and show that our proposed model can learn meaningful encodings.
  arXiv  Detail & Related papers  (2024-10-11T19:18:47Z)
• Adaptive Large Language Models By Layerwise Attention Shortcuts [46.76681147411957]
  LLM-like setups allow the final layer to attend to all of the intermediate layers as it deems fit through the attention mechanism. We showcase four different datasets, namely acoustic tokens, natural language, and symbolic music, and we achieve superior performance for GPT-like architecture.
  arXiv  Detail & Related papers  (2024-09-17T03:46:01Z)
• How Deep Networks Learn Sparse and Hierarchical Data: the Sparse Random Hierarchy Model [4.215221129670858]
  We show that by introducing sparsity to generative hierarchical models of data, the task acquires insensitivity to spatial transformations that are discrete versions of smooth transformations. We quantify how the sample complexity of CNNs learning the SRHM depends on both the sparsity and hierarchical structure of the task.
  arXiv  Detail & Related papers  (2024-04-16T17:01:27Z)
• On Characterizing the Evolution of Embedding Space of Neural Networks using Algebraic Topology [9.537910170141467]
  We study how the topology of feature embedding space changes as it passes through the layers of a well-trained deep neural network (DNN) through Betti numbers. We demonstrate that as depth increases, a topologically complicated dataset is transformed into a simple one, resulting in Betti numbers attaining their lowest possible value.
  arXiv  Detail & Related papers  (2023-11-08T10:45:12Z)
• Implant Global and Local Hierarchy Information to Sequence based Code Representation Models [25.776540440893257]
  We analyze how the complete hierarchical structure influences the tokens in code sequences and abstract this influence as a property of code tokens called hierarchical embedding. We propose the Hierarchy Transformer (HiT), a simple but effective sequence model to incorporate the complete hierarchical embeddings of source code into a Transformer model.
  arXiv  Detail & Related papers  (2023-03-14T12:01:39Z)
• Decomposing neural networks as mappings of correlation functions [57.52754806616669]
  We study the mapping between probability distributions implemented by a deep feed-forward network. We identify essential statistics in the data, as well as different information representations that can be used by neural networks.
  arXiv  Detail & Related papers  (2022-02-10T09:30:31Z)
• Augmenting Convolutional networks with attention-based aggregation [55.97184767391253]
  We show how to augment any convolutional network with an attention-based global map to achieve non-local reasoning. We plug this learned aggregation layer with a simplistic patch-based convolutional network parametrized by 2 parameters (width and depth) It yields surprisingly competitive trade-offs between accuracy and complexity, in particular in terms of memory consumption.
  arXiv  Detail & Related papers  (2021-12-27T14:05:41Z)
• Is Disentanglement enough? On Latent Representations for Controllable Music Generation [78.8942067357231]
  In the absence of a strong generative decoder, disentanglement does not necessarily imply controllability. The structure of the latent space with respect to the VAE-decoder plays an important role in boosting the ability of a generative model to manipulate different attributes.
  arXiv  Detail & Related papers  (2021-08-01T18:37:43Z)
• Redefining Neural Architecture Search of Heterogeneous Multi-Network Models by Characterizing Variation Operators and Model Components [71.03032589756434]
  We investigate the effect of different variation operators in a complex domain, that of multi-network heterogeneous neural models. We characterize both the variation operators, according to their effect on the complexity and performance of the model; and the models, relying on diverse metrics which estimate the quality of the different parts composing it.
  arXiv  Detail & Related papers  (2021-06-16T17:12:26Z)
• The Heterogeneity Hypothesis: Finding Layer-Wise Differentiated Network Architectures [179.66117325866585]
  We investigate a design space that is usually overlooked, i.e. adjusting the channel configurations of predefined networks. We find that this adjustment can be achieved by shrinking widened baseline networks and leads to superior performance. Experiments are conducted on various networks and datasets for image classification, visual tracking and image restoration.
  arXiv  Detail & Related papers  (2020-06-29T17:59:26Z)
• Ensemble Wrapper Subsampling for Deep Modulation Classification [70.91089216571035]
  Subsampling of received wireless signals is important for relaxing hardware requirements as well as the computational cost of signal processing algorithms. We propose a subsampling technique to facilitate the use of deep learning for automatic modulation classification in wireless communication systems.
  arXiv  Detail & Related papers  (2020-05-10T06:11:13Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.