Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations

• URL: http://arxiv.org/abs/2406.09366v1
• Date: Thu, 13 Jun 2024 17:49:56 GMT
• Title: Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations
• Authors: Rylan Schaeffer, Victor Lecomte, Dhruv Bhandarkar Pai, Andres Carranza, Berivan Isik, Alyssa Unell, Mikail Khona, Thomas Yerxa, Yann LeCun, SueYeon Chung, Andrey Gromov, Ravid Shwartz-Ziv, Sanmi Koyejo,
• Abstract summary: Maximum Manifold Capacity Representations (MMCR) is a self-supervised learning method that matches or surpasses other leading MVSSL methods. We show that MMCR incentivizes alignment and uniformity of learned embeddings. We experimentally confirm non-monotonic changes in the pretraining loss akin to double descent. We then show that MMCR, originally applied to image data, is performant on multimodal image-text data.
• Score: 29.69549286641418
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Maximum Manifold Capacity Representations (MMCR) is a recent multi-view self-supervised learning (MVSSL) method that matches or surpasses other leading MVSSL methods. MMCR is intriguing because it does not fit neatly into any of the commonplace MVSSL lineages, instead originating from a statistical mechanical perspective on the linear separability of data manifolds. In this paper, we seek to improve our understanding and our utilization of MMCR. To better understand MMCR, we leverage tools from high dimensional probability to demonstrate that MMCR incentivizes alignment and uniformity of learned embeddings. We then leverage tools from information theory to show that such embeddings maximize a well-known lower bound on mutual information between views, thereby connecting the geometric perspective of MMCR to the information-theoretic perspective commonly discussed in MVSSL. To better utilize MMCR, we mathematically predict and experimentally confirm non-monotonic changes in the pretraining loss akin to double descent but with respect to atypical hyperparameters. We also discover compute scaling laws that enable predicting the pretraining loss as a function of gradients steps, batch size, embedding dimension and number of views. We then show that MMCR, originally applied to image data, is performant on multimodal image-text data. By more deeply understanding the theoretical and empirical behavior of MMCR, our work reveals insights on improving MVSSL methods.

Related papers

• Language Models as Zero-shot Lossless Gradient Compressors: Towards General Neural Parameter Prior Models [66.1595537904019]
  Large language models (LLMs) can act as gradient priors in a zero-shot setting. We introduce LM-GC, a novel method that integrates LLMs with arithmetic coding.
  arXiv  Detail & Related papers  (2024-09-26T13:38:33Z)
• F-LMM: Grounding Frozen Large Multimodal Models [53.8059045627934]
  We present F-LMM -- grounding frozen off-the-shelf LMMs in human-AI conversations. Using only a few trainable CNN layers, we can translate word-pixel attention weights to mask logits. Our F-LMM neither learns special segmentation tokens nor utilises high-quality grounded instruction-tuning data.
  arXiv  Detail & Related papers  (2024-06-09T15:14:26Z)
• Maximum Manifold Capacity Representations in State Representation Learning [8.938418994111716]
  manifold-based self-supervised learning (SSL) builds on the manifold hypothesis. DeepInfomax with an unbalanced atlas (DIM-UA) has emerged as a powerful tool. MMCR presents a new frontier for SSL by optimizing class separability via manifold compression. We present an innovative integration of MMCR into existing SSL methods, incorporating a discerning regularization strategy.
  arXiv  Detail & Related papers  (2024-05-22T17:19:30Z)
• Learning Efficient Coding of Natural Images with Maximum Manifold Capacity Representations [4.666056064419346]
  The efficient coding hypothesis proposes that the response properties of sensory systems are adapted to the statistics of their inputs. While elegant, information theoretic properties are notoriously difficult to measure in practical settings or to employ as objective functions in optimization. Here we outline the assumptions that allow manifold capacity to be optimized directly, yielding Maximum Manifold Capacity Representations (MMCR)
  arXiv  Detail & Related papers  (2023-03-06T17:26:30Z)
• Understanding Multimodal Contrastive Learning and Incorporating Unpaired Data [19.72282903349282]
  We show a general class of nonlinear loss functions for multimodal contrastive learning (MMCL) We quantitatively show that the feature learning ability of MMCL can be better than that of unimodal contrastive learning applied to each modality. When we have access to additional unpaired data, we propose a new MMCL loss that incorporates additional unpaired datasets.
  arXiv  Detail & Related papers  (2023-02-13T10:11:05Z)
• From Cloze to Comprehension: Retrofitting Pre-trained Masked Language Model to Pre-trained Machine Reader [130.45769668885487]
  Pre-trained Machine Reader (PMR) is a novel method for retrofitting masked language models (MLMs) to pre-trained machine reading comprehension (MRC) models without acquiring labeled data. To build the proposed PMR, we constructed a large volume of general-purpose and high-quality MRC-style training data. PMR has the potential to serve as a unified model for tackling various extraction and classification tasks in the MRC formulation.
  arXiv  Detail & Related papers  (2022-12-09T10:21:56Z)
• Provable Generalization of Overparameterized Meta-learning Trained with SGD [62.892930625034374]
  We study the generalization of a widely used meta-learning approach, Model-Agnostic Meta-Learning (MAML) We provide both upper and lower bounds for the excess risk of MAML, which captures how SGD dynamics affect these generalization bounds. Our theoretical findings are further validated by experiments.
  arXiv  Detail & Related papers  (2022-06-18T07:22:57Z)
• MAML is a Noisy Contrastive Learner [72.04430033118426]
  Model-agnostic meta-learning (MAML) is one of the most popular and widely-adopted meta-learning algorithms nowadays. We provide a new perspective to the working mechanism of MAML and discover that: MAML is analogous to a meta-learner using a supervised contrastive objective function. We propose a simple but effective technique, zeroing trick, to alleviate such interference.
  arXiv  Detail & Related papers  (2021-06-29T12:52:26Z)
• Multi-document Summarization with Maximal Marginal Relevance-guided Reinforcement Learning [54.446686397551275]
  We present RL-MMR, which unifies advanced neural SDS methods and statistical measures used in classical MDS. RL-MMR casts MMR guidance on fewer promising candidates, which restrains the search space and thus leads to better representation learning.
  arXiv  Detail & Related papers  (2020-09-30T21:50:46Z)

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.