6,099 Hits in 4.4 sec

Short-term memory ability of reservoir-based temporal difference learning model

Yu Yoshino, Yuichi Katori
2022 Nonlinear Theory and Its Applications IEICE  
A network model with temporal difference (TD) learning and reservoir computing (RC) has been proposed to control autonomous robots.  ...  In the present study, we evaluate the model with a task requiring short-term memory and clarify the reservoir's role in memorizing task-relevant sensory information.  ...  Acknowledgments This paper is based on results obtained from a project, JPNP16007, commissioned by the New Energy and Industrial Technology Development Organization (NEDO), and this work is supported by  ... 
doi:10.1587/nolta.13.203 fatcat:licjqibxofe3fe7lqjh6odadl4

On Tiny Episodic Memories in Continual Learning [article]

Arslan Chaudhry, Marcus Rohrbach, Mohamed Elhoseiny, Thalaiyasingam Ajanthan, Puneet K. Dokania, Philip H. S. Torr, Marc'Aurelio Ranzato
2019 arXiv   pre-print
In this work, we empirically analyze the effectiveness of a very small episodic memory in a CL setup where each training example is only seen once.  ...  the episodic memory, significantly outperforms specifically designed CL approaches with and without episodic memory.  ...  We would also like to acknowledge the Royal Academy of Engineering and FiveAI.  ... 
arXiv:1902.10486v4 fatcat:4iqp24vyanhlvhvsxmfpwv75gm

Neural Kernels for Recursive Support Vector Regression as a Model for Episodic Memory [article]

Christian Leibold
2022 bioRxiv   pre-print
AbstractRetrieval of episodic memories requires intrinsic reactivation of neuronal activity patterns. The content of the memories are thereby assumed to be stored in synaptic connections.  ...  This paper proposes a theory in which these are the synaptic connections that specifically convey the temporal order information contained in the sequences of a neuronal reservoir to the sensory-motor  ...  Posthoc increase of a could thus be considered as a model of memory consolidation, posthoc decrease of a as a model of extiction learning.  ... 
doi:10.1101/2022.02.22.481458 fatcat:xbjpiwwuyvhs7fnnvd3ydn7hhi

Deep Q-network using reservoir computing with multi-layered readout [article]

Toshitaka Matsuki
2022 arXiv   pre-print
An approach with replay memory introducing reservoir computing has been proposed, which trains an agent without BPTT and avoids these issues.  ...  The basic idea of this approach is that observations from the environment are input to the reservoir network, and both the observation and the reservoir output are stored in the memory.  ...  This work was supported by JSPS KAKENHI (Grant-in-Aid for Encouragement of Scientists) Number 21H04323.  ... 
arXiv:2203.01465v1 fatcat:hl5s5u2ihbczro7n3agbbu3aay

Reservoirs learn to learn [article]

Anand Subramoney and Franz Scherr and Wolfgang Maass
2019 arXiv   pre-print
In order to examine the benefits of choosing recurrent weights within a liquid with a purpose, we applied the Learning-to-Learn (L2L) paradigm to our model: We optimized the weights of the recurrent connections  ...  So far only the weights of a linear readout were adapted for a specific task.  ...  We gratefully acknowledge Sandra Diaz, Alexander Peyser and Wouter Klijn from the Simulation Laboratory Neuroscience of the Jülich Supercomputing Centre for their support.  ... 
arXiv:1909.07486v2 fatcat:ldusa77ia5bfhas4e2gfayx3kq

One-shot learning of paired associations by a reservoir computing model with Hebbian plasticity [article]

M Ganesh Kumar, Cheston Tan, Camilo Libedinsky, Shih-Cheng Yen, Andrew Yong-Yi Tan
2021 arXiv   pre-print
Here we extend the model by replacing the symbolic mechanism with a reservoir of recurrently connected neurons resembling cortical microcircuitry.  ...  As with rodents, the reservoir model exhibited one-shot learning for multiple paired associations.  ...  Acknowledgments and Disclosure of Funding  ... 
arXiv:2106.03580v1 fatcat:kr565v3ykjbp3mqoxam7gnjxgi

Reservoir Computing for Detection of Steady State in Performance Tests of Compressors [article]

Eric Aislan Antonelo and Carlos Alberto Flesch and Filipe Schmitz
2017 arXiv   pre-print
This assigned cluster defines a particular binary input that shifts the operating point of the reservoir to a subspace of trajectories for the duration of the episode.  ...  Specifically, we use Reservoir Computing (RC) networks which simplify the learning of RNNs by speeding up training time and showing convergence to a global optimum.  ...  Bernardo Schwedersky for their support on the understanding of the industrial datasets.  ... 
arXiv:1706.00782v1 fatcat:hzwyivo2nbb27ie323nihwmxf4

Reservoir computing for detection of steady state in performance tests of compressors

Eric Aislan Antonelo, Carlos Alberto Flesch, Filipe Schmitz
2018 Neurocomputing  
This assigned cluster defines a particular binary input that shifts the operating point of the reservoir to a subspace of trajectories for the duration of the episode.  ...  Specifically, we use Reservoir Computing (RC) networks which simplify the learning of RNNs by speeding up training time and showing convergence to a global optimum.  ...  Bernardo Schwedersky for their support on the understanding of the industrial datasets.  ... 
doi:10.1016/j.neucom.2017.09.005 fatcat:aynnl6vlcbeajjfm462tdeyrsm

Saliency Guided Experience Packing for Replay in Continual Learning [article]

Gobinda Saha, Kaushik Roy
2021 arXiv   pre-print
While learning a new task, we replay these memory patches with appropriate zero-padding to remind the model about its past decisions.  ...  One way to enable such learning is to store past experiences in the form of input examples in episodic memory and replay them when learning new tasks.  ...  of six centers in JUMP, a Semiconductor Research Corporation program sponsored by DARPA.  ... 
arXiv:2109.04954v1 fatcat:6jogfexpebcwvolr4d4o33xb2y

Deep Reinforcement Learning for Long Term Hydropower Production Scheduling [article]

Signe Riemer-Sorensen, Gjert H. Rosenlund
2020 arXiv   pre-print
We consider a use-case where the aim is to optimise the yearly revenue given week-by-week inflows to the reservoir and electricity prices.  ...  The presented model is not ready to substitute traditional optimisation tools but demonstrates the complementary potential of reinforcement learning in the data-rich field of hydropower scheduling.  ...  However, several of the parameters will mainly influence convergence rate, so if trained for a large enough number of episodes, the final model performance will be similar.  ... 
arXiv:2012.06312v1 fatcat:3xrkwlxexfbv7cbtmjtupcvzsm

Graph-Based Continual Learning [article]

Binh Tang, David S. Matteson
2021 arXiv   pre-print
Rehearsal approaches alleviate the problem by maintaining and replaying a small episodic memory of previous samples, often implemented as an array of independent memory slots.  ...  Empirical results on several benchmark datasets show that our model consistently outperforms recently proposed baselines for task-free continual learning.  ...  rehearsal approach based on an episodic memory of parameter gradients; (4) ER (Chaudhry et al., 2019), a simple yet competitive experience method based on reservoir sampling; (5) MER, a rehearsal approach  ... 
arXiv:2007.04813v2 fatcat:gk7hyy5plfgijdjhr7cjfs3sga

A Reservoir Model of Explicit Human Intelligence [article]

Eric C. Wong
2022 arXiv   pre-print
We propose that these two innovations, together with pre-existing mechanisms for associative learning, allowed us to develop a conceptually simple associative network that operates like a reservoir of  ...  We describe here a network architecture for the human brain that may support this feature and suggest that two key innovations were the ability to consider an offline model of the world, and the use of  ...  Acknowledgements We would like to thank Richard Buxton, Thomas Liu, Peter Bandettini, and Alan Simmons for many helpful discussions regarding the subject of this paper.  ... 
arXiv:2207.07912v1 fatcat:4jfyo7uwubhivoy6ey4k7yy374

Hierarchical Bayesian reservoir memory

Ali Nouri, Hooman Nikmehr
2009 2009 14th International CSI Computer Conference  
So, the model is called Hierarchical Bayesian Reservoir Memory (HBRM).  ...  The model is a hierarchical Bayesian structure that uses Reservoir Computing methods as the state-of-the-art and the most biological plausible Temporal Sequence Processing method for online and unsupervised  ...  ACKNOWLEDGMENT This research was started in Axaya Research Group at Laboratory of Multi-Agent Systems and Distributed Artificial Intelligence of the Department of Computer Engineering, Bu-Ali Sina University  ... 
doi:10.1109/csicc.2009.5349642 fatcat:wapg7sdwx5a4pbmp2hp2plcenq

A Neuromorphic Model with Delay-based Reservoir for Continuous Ventricular Heartbeat Detection

Xiangpeng Liang, Haobo Li, Aleksandra Vuckovic, John Richard Mercer, Hadi Heidari
2021 IEEE Transactions on Biomedical Engineering  
Such computational modelling boosts memory eciency by simplifying the computing procedure and minimizing the required memory for future wearable devices.  ...  We apply delay-based reservoir computing as the information processing core, along with a novel training and labelling method.  ...  Delay-based reservoir computing for ECG processing A conventional reservoir computing network consists of an input layer, a reservoir and an output layer.  ... 
doi:10.1109/tbme.2021.3129306 pmid:34797760 fatcat:i2jpdg3wcjdk3bj625t7o7scge

Example Based Hebbian Learning may be sufficient to support Human Intelligence [article]

Eric C Wong
2019 bioRxiv   pre-print
In this hypothesis paper we argue that a simple Hebbian learning mechanism, along with reinforcement through well-known neuromodulatory systems, can form the basis of a computational theory of learning  ...  We show that when driven by example behavior Hebbian learning rules can support procedural, episodic and semantic memory.  ...  Simmons for helpful review of this manuscript, and Maxwell Wong for help in coding of the examples.  ... 
doi:10.1101/758375 fatcat:vh5cxpzldjbwlpj5itntymxu2i
« Previous Showing results 1 — 15 out of 6,099 results