Filters








58,111 Hits in 3.7 sec

Efficiently Implementing Episodic Memory [chapter]

Nate Derbinsky, John E. Laird
2009 Lecture Notes in Computer Science  
However, providing efficient storage and retrieval in a task-independent episodic memory presents considerable theoretical and practical challenges.  ...  We explore whether even with intractable asymptotic growth, it is possible to develop efficient algorithms and data structures for episodic memory systems that are practical for real-world tasks.  ...  support effective and efficient episodic memory.  ... 
doi:10.1007/978-3-642-02998-1_29 fatcat:c4wk55hezzdcbfa5mffldhem5m

Memory Models for Improving Tabu Search with Real Continuous Variables

Andrew Connor
2006 2006 Sixth International Conference on Hybrid Intelligent Systems (HIS'06)  
The memory model is based on the transfer of events from episodic memory into generalised rules stored in semantic memory.  ...  This paper proposes that current memory models in use for tabu search algorithms are at best evolving, as opposed to adaptive, and that improvements can be made by considering the nature of human memory  ...  In future work, it is proposed to introduce the concepts of both episodic and semantic memory into the long term memory used in the tabu search implementation.  ... 
doi:10.1109/his.2006.264910 fatcat:dqspd4xdanb45czndeutw4t5qq

Towards sample-efficient policy learning with DAC-ML

Ismael T. Freire, Adrián F. Amil, Vasiliki Vouloutsi, Paul F.M.J. Verschure
2021 Procedia Computer Science  
Recent studies have tried to overcome this limitation by adding memory systems and architectural biases to improve learning speed, such as in Episodic Reinforcement Learning.  ...  Recent studies have tried to overcome this limitation by adding memory systems and architectural biases to improve learning speed, such as in Episodic Reinforcement Learning.  ...  Reactive layer implements a pseudo-random-walk generator that allows an efficient random exploration of the environment.  ... 
doi:10.1016/j.procs.2021.06.035 fatcat:p3kinb2hn5bvndx5cpl5t65bfq

A Multi-Domain Evaluation of Scaling in a General Episodic Memory

Nate Derbinsky, Justin Li, John Laird
2021 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
In this paper, we review the implementation of episodic memory in Soar and present an expansive evaluation of that system.  ...  Episodic memory endows agents with numerous general cognitive capabilities, such as action modeling and virtual sensing.  ...  These cues implemented virtual-sensing, detecting-repetition, and action-modeling capabilities.  ... 
doi:10.1609/aaai.v26i1.8151 fatcat:ezbf2rs4ufazvet4prbsdb5rsm

A Cognitive Model of Episodic Memory Integrated with a General Cognitive Architecture

Andrew Nuxoll, John E. Laird
2004 International Conference on Cognitive Modelling  
We present two implementations of a computational model of episodic memory in Soar. We demonstrate all four stages of the model for a simple interactive task.  ...  In this paper, we present a framework for describing the functional stages for computational models of episodic memory: encoding, storage, retrieval and use of the retrieved memories.  ...  However, the memory did not address issues of how to search and retrieve episodes effectively and efficiently.  ... 
dblp:conf/iccm/NuxollL04 fatcat:lzws7orklbfzffnnsponx5ma4i

Comprehensive Working Memory Activation in Soar

John E. Laird, Andrew Nuxoll
2004 International Conference on Cognitive Modelling  
In this paper, we present a comprehensive implementation of working memory activation in Soar that takes advantage of the unique characteristic of Soar's working memory structure, namely persistence.  ...  Memory activation has been modeled in symbolic architectures in the past, but usually at the level of individual chunks or productions in long term memory.  ...  a more efficient implementation, and an implementation that takes advantage of the unique structure of Soar's working memory and distinguishes it from ACT-R's activation scheme.  ... 
dblp:conf/iccm/LairdN04 fatcat:ogfqlxdwive3rpxwwvhuns7oku

Rerun: Exploiting Episodes for Lightweight Memory Race Recording

Derek R. Hower, Mark D. Hill
2008 2008 International Symposium on Computer Architecture  
While sources of nondeterminism in a uniprocessor can be recorded efficiently in software, it seems likely that hardware support will be needed in a multiprocessor environment where the outcome of memory  ...  In particular, Rerun passively creates atomic episodes. Each episode is a dynamic instruction sequence that a thread happens to execute without interacting with other threads.  ...  Conclusions We develop Rerun, a memory race recorder that uses episodes to efficiently log memory reference order.  ... 
doi:10.1109/isca.2008.26 dblp:conf/isca/HowerH08 fatcat:cl3joqnzyfhanjhs64fu4rcyvu

Rerun

Derek R. Hower, Mark D. Hill
2008 SIGARCH Computer Architecture News  
While sources of nondeterminism in a uniprocessor can be recorded efficiently in software, it seems likely that hardware support will be needed in a multiprocessor environment where the outcome of memory  ...  In particular, Rerun passively creates atomic episodes. Each episode is a dynamic instruction sequence that a thread happens to execute without interacting with other threads.  ...  Conclusions We develop Rerun, a memory race recorder that uses episodes to efficiently log memory reference order.  ... 
doi:10.1145/1394608.1382144 fatcat:q5hpmep2hve4bp5zav2mqtzwam

An In-Memory Computing SRAM Macro for Memory-Augmented Neural Network

Sunghoon Kim, Wonjae Lee, Sundo Kim, Sungjin Park, Dongsuk Jeon
2021 IEEE Transactions on Circuits and Systems - II - Express Briefs  
We first propose algorithmic optimizations for a few-shot learning algorithm employing MANN for efficient hardware implementation.  ...  In this paper, we present an SRAM macro designed for accelerating Memory-Augmented Neural Network (MANN).  ...  Our design implements the algorithm in [14] as a baseline, but we further optimize the algorithm for efficient hardware implementation.  ... 
doi:10.1109/tcsii.2021.3132063 fatcat:vomaj2mkareerpycilsoumixri

An Energy Efficient EdgeAI Autoencoder Accelerator for Reinforcement Learning

Nitheesh Kumar Manjunath, Aidin Shiri, Morteza Hosseini, Bharat Prakash, Nicholas R. Waytowich, Tinoosh Mohsenin
2021 IEEE Open Journal of Circuits and Systems  
The hardware is configurable to reach an efficiency of over 1 TOP/J on FPGA implementation.  ...  We designed and implemented a scalable hardware accelerator which is configurable in terms of the number of processing elements (PEs) and memory data width to achieve the best power, performance, and energy  ...  autoencoder hardware that allows experimenting with the number of PEs and memory data widths to achieve the best power, performance, and energy efficiency trade-off for EdgeAI embedded devices. • Implement  ... 
doi:10.1109/ojcas.2020.3043737 fatcat:kzzb2febargvha7ojdazhh76oi

Towards Long-Term Memory for Social Robots: Proposing a New Challenge for the RoboCup@Home League [article]

Matías Pavez, Javier Ruiz del Solar, Victoria Amo, Felix Meyer zu Driehausen
2018 arXiv   pre-print
Long-term memory is essential to feel like a continuous being, and to be able to interact/communicate coherently.  ...  Social robots need long-term memories in order to establish long-term relationships with humans and other robots, and do not act just for the moment.  ...  However, in contrast to implementing semantic memories, the implementation of episodic memories for social robots is much less developed, and therefore we focus our work on the latter.  ... 
arXiv:1811.10758v1 fatcat:vmhaxybjvzgbjcasuldrzhuf6q

A Study of Performance Portability Using Piecewise-Parabolic Method (PPM) Gas Dynamics Applications

Pei-Hung Lin, Jagan Jayaraj, Paul Woodward, Pen-Chung Yew
2012 Procedia Computer Science  
These have included multiple cores per CPU, multiple simultaneous threads per core, and, especially with GPUs, highly complex memory hierarchies.  ...  Each episode is like a subroutine and is called from a main computational kernel. In our GPU implementation, we only generate few GPU kernels, each performing a series of episodes of computation.  ...  First, to have efficient usage of the precious on-chip memory space, we must reduce the memory footprint and reuse cache-resident data.  ... 
doi:10.1016/j.procs.2012.04.217 fatcat:jjrxwbzjwjbblch7anyjodytsu

Enhancing intelligent agents with episodic memory

Andrew M. Nuxoll, John E. Laird
2012 Cognitive Systems Research  
Because of its peripheral nature, the effectiveness, efficiency and completeness of their agent's episodic memory implementation was not investigated.  ...  • How does semantic memory increase the efficiency of episodic memory cueing and retrieval?  ...  Any implementation of recursive retrieval must avoid the danger of infinite recursion. It must also be able to distinguish a "root" memory from a "meta-memory."  ... 
doi:10.1016/j.cogsys.2011.10.002 fatcat:qsbxe4hi7vcb5lffnxakj3nwcy

Machines Learning - Towards a New Synthetic Autobiographical Memory [chapter]

Mathew H. Evans, Charles W. Fox, Tony J. Prescott
2014 Lecture Notes in Computer Science  
This position paper highlights desiderata for a modern implementation of synthetic autobiographical memory based on human episodic memory, and proposes that a recently developed model of hippocampal memory  ...  Autobiographical memory is the organisation of episodes and contextual information from an individual's experiences into a coherent narrative, which is key to our sense of self.  ...  We close with a discussion of our specific implementation goals, centring our episodic memory system within a rich sensorimotor system to aid ongoing processing.  ... 
doi:10.1007/978-3-319-09435-9_8 fatcat:edyzhniozzgcjjfvx65twbnebq

Tiny Eats: Eating Detection on a Microcontroller [article]

Maria T. Nyamukuru, Kofi M. Odame
2020 arXiv   pre-print
This paper describes the implementation of the Tiny Eats GRU, a shallow GRU neural network, on a low power micro-controller, Arm Cortex M0+, to classify eating episodes.  ...  The Tiny Eats GRU utilizes only 4% of the Arm Cortex M0+ memory and identifies eating or non-eating episodes with 6 ms latency and accuracy of 95.15%.  ...  eating episodes that can be implemented directly on a wearable device.  ... 
arXiv:2003.06699v1 fatcat:gc634clqljeazhwrjdp44n6zna
« Previous Showing results 1 — 15 out of 58,111 results