Memory Transformer

2020/06/20 by Mikhail Burtsev, Yuri Kuratov, Burtsev, Mikhail S. +5 · 12 citations
Engineering · #Advanced Memory and Neural Computing #Analog and Mixed-Signal Circuit Design #Computation and Language (cs.CL) #FOS: Computer and information sciences #Ferroelectric and Negative Capacitance Devices #Machine Learning (cs.LG) #Neural and Evolutionary Computing (cs.NE)

paper · pdf · doi:10.48550/arxiv.2006.11527

openalex publication_date 2020/06/20 · openalex created_date 2024/04/11 · openalex updated_date 2026/07/28

Abstract

Transformer-based models have achieved state-of-the-art results in many natural language processing tasks. The self-attention architecture allows transformer to combine information from all elements of a sequence into context-aware representations. However, information about the context is stored mostly in the same element-wise representations. This might limit the processing of properties related to the sequence as a whole more difficult. Adding trainable memory to selectively store local as well as global representations of a sequence is a promising direction to improve the Transformer model. Memory-augmented neural networks (MANNs) extend traditional neural architectures with general-purpose memory for representations. MANNs have demonstrated the capability to learn simple algorithms like Copy or Reverse and can be successfully trained via backpropagation on diverse tasks from question answering to language modeling outperforming RNNs and LSTMs of comparable complexity. In this work, we propose and study few extensions of the Transformer baseline (1) by adding memory tokens to store non-local representations, (2) creating memory bottleneck for the global information, (3) controlling memory update with dedicated layer. We evaluate these memory augmented Transformers and demonstrate that presence of memory positively correlates with the model performance for machine translation and language modelling tasks. Augmentation of pre-trained masked language model with memory tokens shows mixed results for tasks from GLUE benchmark. Visualization of attention patterns over the memory suggest that it improves the model's ability to process a global context.

Cited by

• Unboxing the Black Box: Mechanistic Interpretability for Algorithmic Understanding of Neural Networks
• Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era
• Catch Your Breath: Adaptive Computation for Self-Paced Sequence Production
• Mem-α: Learning Memory Construction via Reinforcement Learning
• KiVi: Kinesthetic-Visuospatial Integration for Dynamic and Safe Egocentric Legged Locomotion
• Enhancing Transformer-Based Vision Models: Addressing Feature Map Anomalies Through Novel Optimization Strategies
• Extending Foundational Monocular Depth Estimators to Fisheye Cameras with Calibration Tokens
• MIRIX: Multi-Agent Memory System for LLM-Based Agents
• A standard transformer and attention with linear biases for molecular conformer generation
• Learning to Insert [PAUSE] Tokens for Better Reasoning
• Pause Tokens Strictly Increase the Expressivity of Constant-Depth Transformers
• Vision Transformers with Self-Distilled Registers

Related

• Augmenting Self-attention with Persistent Memory
• HMT: Hierarchical Memory Transformer for Efficient Long Context Language Processing
• An Evolved Universal Transformer Memory
• Scaling Transformer to 1M tokens and beyond with RMT
• MemoryPrompt: A Light Wrapper to Improve Context Tracking in Pre-trained Language Models