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 novel memory mechanism for reinforcement learning agents that stores past events in human-readable language.
problem Lack of interpretability in reinforcement learning agent's memory mechanisms.
method Uses CLIP to associate visual inputs with language tokens, then feeds these tokens to a pretrained language model.
result Significantly faster convergence on challenging continuous recognition tasks.
The paper studies scaling laws for associative memory mechanisms.
problem Understanding and optimizing learning and memorization processes.
method High-dimensional matrices of outer products of embeddings, relating to transformer models. Derived scaling laws with sample and parameter sizes. Extensive numerical experiments.
result Precise scaling laws and statistical efficiency of estimators.
Deep neural networks can store and recall data efficiently.
problem Identifying computational mechanisms for memorization and retrieval of data.
method Training overparameterized autoencoders and sequence encoders using standard optimization methods.
result Overparameterized autoencoders and sequence encoders store and recall data efficiently as attractors.
Paper shows how LSTM can remember long sequences by attending to persisted information.
problem LSTMs struggle with long sequences due to fading information and bias towards recent data.
method The paper introduces a mechanism that allows LSTMs to attend to information in memory based on how long it was persisted by the gating mechanism.
result The method improves LSTM's ability to process long sequences by retrieving information proportionally to its persistence in memory.
A new algorithm uses reservoir sampling to enhance a reinforcement learning agent's memory.
problem Efficiently maintaining and recalling past states for reinforcement learning.
method Reservoir sampling to maintain a fixed number of past states for an external memory.
result The method allows for efficient online computation of gradient estimates.
AUTR generates sentences using a dynamic memory and attention mechanism.
problem Generating coherent sentences without explicit training data.
method Recurrent neural network with dynamic attention and canvas memory.
result AUTR learns meaningful latent representations and achieves competitive performance.
New memory models improve LSSVM's generalization and reduce time cost.
problem Improving LSSVM's generalization performance and reducing training time.
method Introducing MIMM and WIMM models that modify LSSVM's memory influence mechanism.
result MIMM and WIMM models outperform LSSVM in generalization and time efficiency.
SMT improves robotic long-horizon tasks by embedding and utilizing past observations.
problem Long-horizon tasks in partially observable environments require effective long-term memory.
method Scene Memory Transformer (SMT) embeds and uses attention to exploit spatio-temporal dependencies.
result SMT outperforms existing policies in visual navigation tasks.
MEMGAN uses memory to improve anomaly detection by isolating abnormal data.
problem Weak guarantees for detecting anomalous data in classical algorithms.
method Memory-augmented Generative Adversarial Networks (MEMGAN) with a memory module.
result MEMGAN provides strong guarantees for anomaly detection with improved reconstruction.
A new model for sequential memory using temporal predictive coding.
problem Forming accurate memory of sequential stimuli in the brain.
method Proposes a novel PC-based model called temporal predictive coding (tPC).
result Shows that tPC models can accurately memorize and retrieve sequential inputs.
Hierarchical memory network tackles large-scale memory access with reduced computation and training complexity.
problem Efficiently accessing large memories in neural networks.
method Hybrid approach combining soft and hard attention mechanisms, using Maximum Inner Product Search (MIPS).
result Hierarchical memory network achieves faster and more trainable memory access than flat memory structures.
Paper explores a new model for non-Markovian control learning.
problem Sequential control learning with delayed rewards in partially observable environments.
method Gated End-to-End Memory Policy Network (GEMPN) using value-based algorithms.
result Encouraging results show the model's capability in non-Markovian decision processes.
Models predict emotional valence from narratives, matching human raters.
problem Predicting emotional valence from multimodal time-series data.
method Adapted attention-based mechanisms (Transformer, Memory Fusion Network) to emotional narratives.
result Models perform well, matching human raters on emotional valence prediction.
RUM improves RNN's long-term memory by using unitary matrices.
problem Limited capacity of RNN to manipulate long-term memory.
method Proposes Rotational Unit of Memory (RUM) with unitary matrices.
result RUM learns long-term dependencies and improves state-of-the-art results.
A new memory replay mechanism improves reinforcement learning stability and speed.
problem Forgetting in reinforcement learning with continuous control.
method Augmented Memory Replay (AMR) that optimizes the replay of past experiences.
result AMR enhances stability and convergence speed of learning algorithms.
Proposes a plastic neural memory model for better anomaly detection.
problem Static attention mechanisms limit NMNs in anomaly detection.
method Introduces dynamic connection weights for improved knowledge retrieval.
result Outperforms state-of-the-art in three medical anomaly detection tasks.
LSTM-FCN improves time series classification with minimal model size increase.
problem Classifying time series sequences with high accuracy.
method Augmented fully convolutional networks with LSTM sub-modules and attention mechanism.
result LSTM-FCN achieves state-of-the-art performance.
A new model MTNet uses memory networks for better time series forecasting.
problem Challenges in modeling complex patterns and dependencies in multivariate time series data.
method Memory Time-series network (MTNet) with a large memory component, three encoders, and an autoregressive component.
result MTNet effectively captures long-term patterns and incorporates information from other variables.
Metalearned neural memory improves learning across various tasks.
problem Improving neural network adaptability and memory function.
method Augmenting neural networks with a metalearned external memory mechanism.
result The model achieves strong performance on diverse learning problems.
A new recurrent unit alleviates vanishing gradients for long-term dependencies.
problem Vanishing gradients in recurrent neural networks make long-term dependencies hard to model.
method Proposes a new NRU architecture that avoids saturating activation functions and gates.
result Demonstrates superior performance across various tasks with and without long-term dependencies.
Improved RNNs reduce memory decay and enhance language tasks.
problem Memory decay in RNNs affects performance in sequence prediction tasks.
method Introduced trainable scaling factors and a dependent bidirectional RNN to mitigate memory decay and improve performance.
result The proposed ELSTM and DBRNN models achieved up to 30% improvement in LAS compared to LSTM and GRU in dependency parsing.
Memory-Augmented Recurrent Networks improve dialogue coherence by expanding conversation history storage.
problem Fixed-size vectors limit dialogue coherence; attention mechanisms are computationally expensive.
method Introduce Neural Turing Machines (NTMs) to provide flexible and permanent storage for dialogue history.
result Improved perplexity performance compared to existing baselines.
A caching mechanism improves sequence to logical form generation accuracy.
problem Generating logical forms from natural language sequences.
method Proposes a caching mechanism to increase output probability of source input tokens and weigh them based on context.
result Improves sequence/token-level accuracy on sequence to logical form tasks.
A new memory system learns and generates like new data.
problem Training generative models on new data.
method Hierarchical conditional generative model with distributed memory.
result The memory system significantly improves generative models.
V-HMN integrates memory mechanisms for improved image recognition.
problem Limited interpretability and high data requirements of existing vision backbones.
method Brain-inspired hierarchical memory modules with iterative refinement.
result V-HMN achieves strong performance on image classification benchmarks.
Study combines CNNs and LSTMs for ECG classification, improving performance with attention mechanisms.
problem Improving ECG classification accuracy for weakly supervised tasks.
method Investigates ConvLSTM networks and various attention mechanisms to capture both morphological and temporal features.
result Attention mechanisms improve classification performance and provide interpretable saliency maps.
A new memory model enhances deep learning's visual understanding.
problem Lack of short-term memory in deep learning models.
method Introduces a biologically inspired visual working memory architecture.
result Model achieves competitive classification performance and reconstructs images.
Service-induced congestion in memory-constrained LLM serving
problem Service-induced congestion in memory-constrained large language model (LLM) serving
method Developing a discrete-time dynamical model of memory-constrained LLM inference
result The system converges to a unique worst-case limit cycle that is asymptotically stable outside a Lebesgue-measure-zero exact-capture set, with throughput losses as large as 50%.
Hydra boosts efficiency for long-context reasoning in resource-constrained settings.
problem Quadratic complexity of transformers limits long-context reasoning in resource-constrained systems.
method Hydra uses a modular architecture with adaptive routing between sparse global attention, mixture-of-experts, and dual memories.
result Hydra achieves significant throughput and accuracy improvements for long-context reasoning.
Novel memory access mechanism improves complex reasoning tasks.
problem Challenges in multi-fact question-answering and positional reasoning.
method Gated End-to-End Memory Network architecture with a novel access regulation mechanism.
result Significant improvements on challenging tasks in the 20 bAbI dataset and DSTC-2.
New cooperative dynamics enhances retrieval performance in neural networks.
problem Understanding emergent computational capabilities in disordered systems.
method Leveraging statistical mechanics, extended neural network architecture for hetero-associative memory.
result Layers trained with less informative datasets develop retrieval regions of the same amplitude, leading to optimal performance.
ARMIN improves memory efficiency and lightness in neural networks.
problem Complex memory addressing and inefficient memory information exploitation in previous MANNs.
method ARMIN uses hidden state for automatic memory addressing and a novel RNN cell for memory integration.
result ARMIN achieves lower computational overhead and similar performances compared to vanilla LSTM.
Energy Transformer integrates attention, energy models, and associative memory.
problem Lack of clear theoretical foundations in attention mechanisms and straightforward design of energy functions in energy-based models.
method Proposes Energy Transformer, a sequence of attention layers with a specifically engineered energy function.
result Obtained strong results on graph anomaly detection and classification tasks.
This paper improves RNN memory capacity for long sequences through learning associative memory update rules.
problem Challenges in RNNs remembering long sequences.
method Jointly learns memory update rule with task objective and uses multiple associative memories.
result Improves memory capacity for long sequence encoding.
We propose a stochastic process driven by memory effect with novel distributions including both exponential and leptokurtic heavy-tailed distributions. A class of distribution is analytically derived from the continuum limit of the discrete binary process with the renormalized auto-correlation and the closed form momen…
New approach reduces malware detection memory requirements and speeds up training.
problem Efficiently classifying long sequences of malware detection data.
method Developed a new temporal max pooling method and global channel gating design.
result 116x more memory efficient and 25.8x faster training on original dataset.
HiPPO framework optimizes memory compression for sequential data.
problem Incremental representation of cumulative history in sequential data.
method Optimal polynomial projections for online function approximation.
result HiPPO-LegS achieves state-of-the-art accuracy on MNIST.
It is generally accepted that many time series of practical interest exhibit strong dependence, i.e., long memory. For such series, the sample autocorrelations decay slowly and log-log periodogram plots indicate a straight-line relationship. This necessitates a class of models for describing such behavior. A popular cl…
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.
Linear Memory Network separates memory and function in RNNs.
problem Complex transduction problems requiring memory and input-output exploitation.
method Conceptual separation between memory and function, using feedforward and autoencoder components.
result Efficient training and competitive performance on polyphonic music datasets.
This paper proposes a new approach to Transformers by integrating hierarchical associative memory with MetaFormers.
problem Theoretical framework for Transformers and MLP-Mixers remains underdeveloped.
method Integrating hierarchical associative memory with MetaFormers to create a parallelized MLP-Mixer.
result Symmetry-breaking effects improve the performance of the MLP-Mixer in image recognition tasks.
New neural network predicts stock price jumps using limit order book data.
problem Predicting short-term price movements in stock markets.
method Attention-based Convolutional Long Short-Term Memory network architecture.
result Attention mechanism improves jump prediction performance.
A new Hopfield model reduces inefficiency in large transformer models.
problem Inefficiency in training large transformer-based models.
method Introduces an Outlier-Efficient Hopfield Model (OutEffHop) to improve model performance.
result Achieves up to 26% reduction in model output norms across four models.
Extended LSTMs improve volatility prediction by 20%.
problem Predicting asset price volatility with long memory.
method Extended LSTMs with multiple flexible timescales.
result Extended LSTMs outperform rough volatility predictions by 20%.
State-regularized RNNs improve interpretability and performance on long-term memory tasks.
problem RNNs struggle with long-term memory and lack of interpretability.
method Introduce a stochastic state transition mechanism to limit state transitions to a finite set.
result State-regularized RNNs perform better on tasks requiring long-term memory.
New method optimizes memory usage in neural networks, improving sequential learning.
problem Current memory models in neural networks waste memory and computation.
method Formulated an optimization problem to maximize information storage, introduced Cached Uniform Writing.
result Proved Cached Uniform Writing optimizes memory usage and outperforms other methods.
It has been known for a long time that the classical spherical perceptrons can be used as storage memories. Seminal work of Gardner, \cite{Gar88}, started an analytical study of perceptrons storage abilities. Many of the Gardner's predictions obtained through statistical mechanics tools have been rigorously justified. …