New analysis tightens memory capacity of Hopfield models using spherical codes.
problem Optimizing memory capacity in modern Hopfield models and Kernelized Hopfield Models.
method Connecting Hopfield models to spherical codes in information theory, establishing an optimal capacity bound and a sub-linear algorithm.
result First tight and optimal asymptotic memory capacity for modern Hopfield models, matching known lower bounds.
The study evaluates memory and capacity of graph embedding methods.
problem Assessing the memory and capacity of graph embedding methods.
method Not specified in the abstract provided.
result Not specified in the abstract provided.
Generalizes memory and forecasting capacities for nonlinear recurrent networks with dependent inputs.
problem Understanding memory and forecasting capabilities in networks with dependent inputs.
method Formulated bounds for memory and forecasting capacities in terms of network size and input properties.
result Proved that memory capacity for linear recurrent networks with independent inputs is given by the rank of the controllability matrix.
Memory capacity of DAM scales exponentially with feature separation, unaffected by correlations.
problem Understanding how feature correlations impact DAM's capacity.
method Developed an empirical framework to analyze DAM's capacity under varying feature correlations and pattern separations.
result Memory capacity scales exponentially with feature separation, unaffected by correlations.
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.
Study characterizes memory capacity of quantum reservoirs using transmon qubits.
problem Understanding the memory capacity of quantum reservoirs built with transmon qubits.
method Characterized memory capacity of quantum reservoirs using transmon qubits from IBM, focusing on NMSE and topology complexity.
result Found a peak in memory capacity for configurations with n-1 self-loops, suggesting optimal design for forecasting tasks.
Enhances sequence memory capacity in neural networks.
problem Limited sequence capacity in Hopfield-like neural networks.
method Introducing a nonlinear interaction term and a generalized pseudoinverse rule.
result Significantly increased sequence capacity with novel scaling laws.
Enhanced Hopfield model boosts memory retrieval capacity.
problem Memory retrieval in modern Hopfield models with limited capacity.
method Introduces a learnable feature map transforming energy function into kernel space, minimizing separation loss for uniform memory distribution.
result Significant reduction in metastable states, enhancing memory capacity and retrieval accuracy.
New deep ESN architectures improve memory capacity and prediction accuracy.
problem Improving memory capacity and prediction accuracy of ESNs.
method Two new deep ESN architectures: parallel and series. Analysis of memory capacity and prediction accuracy.
result Parallel deep ESNs have equivalent memory capacity to shallow ESNs, while series deep ESNs have smaller memory capacity.
Nonlinear RNNs' memory capacity varies widely, making it impractical.
problem The usefulness of memory capacity as a metric for linear RNNs is questioned.
method Analysis of random nonlinear RNNs with varying input scales.
result Memory capacity of nonlinear RNNs is arbitrary and impractical.
The memory capacity of linear echo state networks is accurately calculated using new numerical methods.
problem Numerical evaluations of memory capacity in recurrent neural networks often contradict theoretical bounds.
method Developed robust numerical approaches exploiting MC neutrality with respect to the input mask matrix.
result Memory curves fully agree with theory when using the proposed methods.
Eye tracking measures ADHD-related working memory deficits.
problem Diagnosing ADHD in adults requires reliable measures of working memory capacity.
method Eye tracking technology and machine learning applied to a working memory task.
result Machine learning generated features unique to ADHD.
A new memory system handles non-stationary environments by self-sizing and retaining memories.
problem Non-stationary environments where memories arrive over time and must be distinguished from noise.
method A self-sizing continual associative memory that generalizes Hopfield's network, handling adaptation and novelty.
result The memory system grows to the intrinsic memory demand of the environment and retains memories without forgetting.
Derives an empirical capacity model for self-attention neural networks.
problem Theoretical capacity of large transformer models is not fully utilized by current optimization algorithms.
method Analyzes memory capacity of transformers using synthetic training data and common training algorithms.
result Derives an empirical capacity model (ECM) for a generic transformer.
Sparse neural encoding can store more memories as targets become sparser.
problem Storing sparse input-target associations in neural networks.
method Mathematical proofs using properties of random polytopes and sub-gaussian random vector variables.
result The capacity of neural maps increases with sparsity in target layers.
Sharp limits found for storing and retrieving input-output associations in linear associative memories.
problem Understanding the fundamental limits of storing and retrieving input-output associations in neural networks.
method Study of a minimal linear associative memory model, introducing a decoupled model and using statistical physics to characterize storage capacity.
result Linear associative memory can store up to 1/2 log(p) associations, providing a sharp statistical-physics characterization.
gLSTM improves graph neural networks by increasing storage capacity to prevent over-squashing.
problem Over-squashing in GNNs collapses information from a large receptive field into a single vector, creating an information bottleneck.
method Introduced a new synthetic task to measure over-squashing and adapted ideas from sequence modeling to develop gLSTM, a novel GNN architecture with improved capacity.
result gLSTM architecture demonstrates strong performance on synthetic and real-world graph benchmarks, mitigating over-squashing.
DMs emerge from DenseAMs, transitioning from memorization to generalization.
problem Hindered memory retrieval in DenseAMs due to spurious states.
method Examined diffusion models through the lens of DenseAMs, focusing on their generative process.
result Identified a critical phase in DMs transitioning from memorization to generalization.
Sparse codes improve optimal control tasks with correlated inputs.
problem Optimal control tasks with correlated feature inputs.
method Used a sparse code to represent natural images in an optimal control task solved with neuro-dynamic programming.
result An over-complete sparse code increases memory capacity and learning speed beyond a complete code.
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. …
New neural networks model complex phenomena with fewer parameters.
problem Challenges in studying higher-order interactions in neural networks.
method Introducing curved neural networks using the maximum entropy principle.
result Curved neural networks accelerate memory retrieval and exhibit explosive phase transitions.
The brain optimizes memory by forgetting what's predictable, improving generalization.
problem Memory consolidation struggles with representational drift, semanticisation, and offline replay.
method Proposes predictive forgetting as a mechanism to optimize generalization by reducing complexity.
result Predictive forgetting improves information-theoretic generalization bounds on stored representations.
Extends DAMs to Gaussian distributions for efficient pattern storage and retrieval.
problem Limited storage capacity and retrieval methods for non-vector pattern representations.
method Introduces a log-sum-exp energy function over Gaussian distributions, using optimal transport maps for retrieval dynamics.
result Proves exponential storage capacity and provides quantitative retrieval guarantees.
New analysis shows ESNs can handle multidimensional inputs without scaling network size.
problem Understanding the memory capacity of ESNs for multidimensional inputs.
method Advanced random matrix theory applied to ESNs with structured inputs.
result Linear scaling of network size with information rate and poly-logarithmic scaling with input dimension.
Study capacity constraints in continual learning with a simple model.
problem Understanding optimal resource allocation for agents with limited memory and compute resources.
method Analyzes a capacity-constrained linear-quadratic-Gaussian (LQG) sequential prediction problem and demonstrates optimal capacity allocation strategies.
result Derives a solution to the capacity-constrained LQG sequential prediction problem and shows how to optimally allocate capacity across sub-problems in the steady state.
Reverse Experience Replay improves Deep Q-learning for sparse rewards.
problem Sparse rewards and reward-maximizing tasks in Deep Q-learning.
method Sampling transitions in reverse order for training.
result Significantly increased performance in tasks with limited experience and memory capacity.
Transformers can store facts efficiently using associative memories.
problem Understanding how transformers store and recall factual information.
method Proved linear scaling of storage capacities for linear and MLP associative memories, introduced a synthetic task, and analyzed gradient flow.
result Shallow transformers can achieve near optimal storage capacity for factual recall tasks using associative memories.
The study revisits Hopfield's associative memory model and calculates its capacity for two specific pattern basins.
problem Determining the capacity of a Hebbian-Hopfield network for storing binary patterns.
method Using fully lifted random duality theory and numerical analysis, the study calculates the capacity for two specific pattern basins.
result Explicit characterizations of the capacity for the AGS and NLT pattern basins, with remarkable fast lifting convergence.
Linear memory stores associations up to a logarithmic scale, but listwise retrieval can handle a quadratic scale.
problem How many key-value associations can a linear memory store?
method Analyzed linear memory models for top-1 and listwise retrieval, proving phase transitions and developing asymptotic theories.
result Linear memory has a logarithmic capacity for top-1 retrieval and a quadratic capacity for listwise retrieval.
New proof shows neural networks can memorize training data with high accuracy.
problem Proving neural networks can memorize training data with high accuracy.
method Probabilistic construction exploiting sparsity.
result Proves neural networks with threshold or mixed threshold-ReLU activations can memorize training data.
A new reservoir computing approach on the hypersphere surpasses memory limits.
problem Sequence learning and time series prediction problems.
method Random fixed-weight RNN on the unit hypersphere, removing non-linear activation.
result Memory capacity exceeds reservoir dimensionality, surpassing typical ESN limits.
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%.
A scalable approach for distributed submodular maximization under fixed machine capacity.
problem Large-scale machine learning problems constrained by limited machine capacity.
method Proposes a truly scalable approach for distributed submodular maximization.
result Achieves performance competitive with centralized greedy solution.
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.
Method interprets LSTMs at the cell level for better understanding of their dynamics.
problem Understanding the dynamics of LSTMs at the cell level.
method A systematic pipeline for interpreting individual hidden state dynamics using response characterization methods.
result Identifies neurons with insightful dynamics and quantifies their impact on network performance.
New model DWM mimics human memory, learning to retain, ignore or forget.
problem Lack of effective separation between episodic and working memory in neural networks.
method Designed Differentiable Working Memory (DWM) inspired by psychological studies.
result DWM learns psychology-inspired tasks faster and generalizes to longer sequences.
Investigates scaling deep neural networks to avoid capacity issues.
problem Avoiding the shattering problem in deep neural networks.
method Formulates conjecture and studies various architectures to determine scaling relations.
result Reveals scaling relations for deep residual networks and recurrent networks.
Unified approach for lifelong learning in recurrent neural networks.
problem Catastrophic forgetting and capacity saturation in lifelong learning.
method Proposed a curriculum-based benchmark and unified model combining Gradient Episodic Memory and Net2Net.
result Unified model performs better than constituent models in lifelong learning setting.
Stable Hadamard Memory improves reinforcement learning by efficiently managing memory.
problem Memory models struggle in partially observable reinforcement learning environments.
method Introduces a novel memory model using the Hadamard product for efficient memory management and updates.
result Significantly outperforms state-of-the-art memory-based methods on challenging benchmarks.
Product Kanerva Machines dynamically combine smaller models for better memory organization.
problem Limited organization in the Kanerva Machine.
method Introducing Product Kanerva Machines that dynamically combine multiple smaller Kanerva Machines.
result Product Kanerva Machines can discover spatial tunings that approximately factorize simple images by object.
Estimates memory usage of reinforcement learning policies by mutual information.
problem Understanding the amount of memory used by reinforcement learning policies.
method Estimates mutual information between behavior histories and current actions.
result Provides a lower bound on minimal memory capacity of any agent.
New framework captures long-term decision dependence in online learning.
problem Long-term dependence on past decisions in online learning.
method Introduces Online Convex Optimization with Unbounded Memory (OCO-UMB) and p p p -effective memory capacity. result Proves O ( H p T ) O(\sqrt{H_p T}) O ( H p T ) upper bound on policy regret and matching lower bound. Muon spectral optimizer outperforms SGD in associative memory tasks.
problem Understanding the advantage of spectral optimizers in learning associative memory.
method Linear associative memory problem, Gaussian inputs and outputs, power law frequency distribution, thresholded gradient approximation.
result Muon significantly outperforms SGD in storage capacity and recovery rates.
Dense Associative Memories outperform classical networks in robustness and signal processing.
problem Improving neural network performance in adversarial attacks and weak signal processing.
method Relaxing replica symmetry in statistical mechanics of spin glasses to analyze unsupervised and supervised learning.
result Explicit analytical investigation of phase diagrams and storage capacities for Dense Associative Memories.
Efficiently trains large models on limited-memory accelerators.
problem Training large-scale machine learning models on limited-memory accelerators.
method Novel algorithmic building block using primal-dual coordinate methods and duality gap information.
result Order-of-magnitude speedup for training generalized linear models on large datasets.
The paper analyzes the performance of delay-based reservoir computing using eigenvalue analysis.
problem Quantifying the performance of delay-based reservoir computing.
method Eigenvalue analysis of the dynamical system to predict reservoir computing performance.
result The performance of a reservoir computing system can be predicted by analyzing the small signal response and eigenvalue spectrum.
LLMs can be tricked into recalling facts based on context clues.
problem Manipulation of LLMs' factual recall through context changes.
method Mathematical exploration of transformers' associative memory properties.
result Transformers use self-attention and value matrix for associative memory.
PipeDream-2BW accelerates large model training by 20x with minimal memory usage.
problem Training large models requires memory beyond single accelerator capacity.
method Pipeline parallelism, weight gradient coalescing, double buffering.
result Accelerates large model training by up to 20x.