arXiv research

Analyzes memory time span in LSTMs for multi-speaker speech separation.

Paper uses LSTM autoencoder for ADS-B data to detect surveillance aircraft.

Adaptive attention span extends Transformer's context size.

FinMem enhances LLM trading agents with layered memory and character design.

Reinforcement learning (RL) agents performing complex tasks must be able to remember observations and actions across sizable time intervals. This is especially true during the initial learning stages, when exploratory behaviour can increase the delay between specific actions and their effects. Many new or popular appro…

The extension of deep learning towards temporal data processing is gaining an increasing research interest. In this paper we investigate the properties of state dynamics developed in successive levels of deep recurrent neural networks (RNNs) in terms of short-term memory abilities. Our results reveal interesting insigh…

We introduce the "NoBackTrack" algorithm to train the parameters of dynamical systems such as recurrent neural networks. This algorithm works in an online, memoryless setting, thus requiring no backpropagation through time, and is scalable, avoiding the large computational and memory cost of maintaining the full gradie…

We investigate the time series of the degree of minimum spanning trees obtained by using a correlation based clustering procedure which is starting from (i) asset return and (ii) volatility time series. The minimum spanning tree is obtained at different times by computing correlation among time series over a time windo…

We study the long memory of order flow for each of three liquid currency pairs on a large electronic trading platform in the foreign exchange (FX) spot market. Due to the extremely high levels of market activity on the platform, and in contrast to existing empirical studies of other markets, our data enables us to perf…

One of the core tasks in multi-view learning is to capture relations among views. For sequential data, the relations not only span across views, but also extend throughout the view length to form long-term intra-view and inter-view interactions. In this paper, we present a new memory augmented neural network model that…

AnEn uses past analogs for weather forecasting, but this work replaces the dataset with deep generative models.

A new method increases 3D medical image segmentation accuracy and speed.

A novel memory network improves interpretability and training for EHR analysis.

Deep learning detects inaccurate smart meters for resource savings.

We propose a new model for the level I of a Limit Order Book (LOB), which incorporates the information about the standing orders at the opposite side of the book after each price change and the arrivals of new orders within the spread. Our main result gives a diffusion approximation for the mid-price process. To illust…

mGRN improves multivariate time series prediction by managing marginal and joint memories.

The Soil Moisture Active Passive (SMAP) mission has delivered valuable sensing of surface soil moisture since 2015. However, it has a short time span and irregular revisit schedule. Utilizing a state-of-the-art time-series deep learning neural network, Long Short-Term Memory (LSTM), we created a system that predicts SM…

Accelerators with power-law memory are proposed in the framework of the discrete time approach. To describe discrete accelerators we use the capital stock adjustment principle, which has been suggested by Matthews.The suggested discrete accelerators with memory describe the economic processes with the power-law memory …

Let $M$ be an $n$-dimensional complete simply connected Riemannian manifold with sectional curvature bounded above by a nonpositive constant $-κ^2$. Using the cone total curvature $TC(Γ)$ of a graph $Γ$ which was introduced by Gulliver and Yamada Math. Z. 2006, we prove that the density at any point of a soap film-like…

New model forecasts long-memory series with time-varying parameters.

Neural network tackles continual learning with neuromodulation and local error signals.

Exploiting sparsity enables hardware systems to run neural networks faster and more energy-efficiently. However, most prior sparsity-centric optimization techniques only accelerate the forward pass of neural networks and usually require an even longer training process with iterative pruning and retraining. We observe t…

Non-spanning identification of scheduled event risk in option pricing.

We study the long-term memory in diverse stock market indices and foreign exchange rates using the Detrended Fluctuation Analysis(DFA). For all daily and high-frequency market data studied, no significant long-term memory property is detected in the return series, while a strong long-term memory property is found in th…

A new hierarchical clustering method selects representative points from sub-minimum-spanning-trees.

New solver avoids memory issues for long differential equations.

This study examines memory effects in S&P500 market correlations using Langevin models.

New memory models improve LSSVM's generalization and reduce time cost.

New method embeds time span into self-attention for better temporal pattern recognition.

This research predicts stock market movements using Vision-Language models.

Correlation matrices of foreign exchange rate time series are investigated for 60 world currencies. Minimal Spanning Tree (MST) graphs for the gold, silver and platinum are presented. Inverse power like scaling is discussed for these graphs as well as for four distinct currency groups (major, liquid, less liquid and no…

A new model MTNet uses memory networks for better time series forecasting.

Stochastic attention learns to retrieve and generate from memory without training.

STanHop predicts multivariate time series with memory-enhanced capabilities.

Optimizes memory and computation in wearable systems.

The paper develops efficient algorithms for sampling from random spanning trees and determinantal point processes.

We study the problem of recovering the subspace spanned by the first $k$ principal components of $d$-dimensional data under the streaming setting, with a memory bound of $O(kd)$. Two families of algorithms are known for this problem. The first family is based on the framework of stochastic gradient descent. Nevertheles…

Delay-SDE-net models time series with memory and uncertainty, outperforming other models.

Extended LSTMs improve volatility prediction by 20%.

Let X and Y be infinite graphs, such that the automorphism group of X is nonamenable, and the automorphism group of Y has an infinite orbit. We prove that there is no automorphism-invariant measure on the set of spanning trees in the direct product X times Y. This implies that the minimal spanning forest corresponding …

TASO optimizes CNN models for memory-constrained devices.

SAGE improves memory efficiency by selectively adding, merging, or ignoring new facts.

New framework tests mean-variance spanning in high dimensions.

This work extends SVM error bounds to weighted SVM and introduces hyperparameter selection methods.

We apply the Zipf power law to financial time series of WIG20 index daily changes (open-close). Thanks to the mapping of time series signal into the sequence of 2k+1 'spin-like' states, where k=0, 1/2, 1, 3/2, ..., we are able to describe any time series increments, with almost arbitrary accuracy, as the one of such 's…

PMI-Masking improves MLM pretraining by masking correlated spans efficiently.

We consider the problem of change-point detection in multivariate time-series. The multivariate distribution of the observations is supposed to follow a graphical model, whose graph and parameters are affected by abrupt changes throughout time. We demonstrate that it is possible to perform exact Bayesian inference when…

Paper presents forecasting models for platelet demand.