Preventing Catastrophic Interference in Multiple-Sequence Learning Using Coupled Reverberating Elman Networks

Ans, Bernard; Rousset, Stéphane; French, Robert M.; Musca, Serban

doi:10.4324/9781315782379-50

Cited by 9 publications

(6 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the same period, French introduced the pseudo-recurrent connectionist network [41,40], a model which makes use of pseudo replay [92] (replay based on random, but fixed, input patterns), but did not address sequential data processing tasks. Later on, the pseudo recurrent network together with pseudo-replay inspired the Reverberating Simple Recurrent Network (RSRN) [9,8]. This is a dual model composed by two auto-associative recurrent networks which exchange information by means of pseudo patterns.…”

Section: Survey Of Continual Learning In Recurrent Modelsmentioning

confidence: 99%

Continual Learning for Recurrent Neural Networks: an Empirical Evaluation

Cossu,

Carta,

Lomonaco

et al. 2021

Preprint

View full text Add to dashboard Cite

Learning continuously during all model lifetime is fundamental to deploy machine learning solutions robust to drifts in the data distribution. Advances in Continual Learning (CL) with recurrent neural networks could pave the way to a large number of applications where incoming data is non stationary, like natural language processing and robotics. However, the existing body of work on the topic is still fragmented, with approaches which are application-specific and whose assessment is based on heterogeneous learning protocols and datasets. In this paper, we organize the literature on CL for sequential data processing by providing a categorization of the contributions and a review of the benchmarks. We propose two new benchmarks for CL with sequential data based on existing datasets, whose characteristics resemble real-world applications. We also provide a broad empirical evaluation of CL and Recurrent Neural Networks in class-incremental scenario, by testing their ability to mitigate forgetting with a number of different strategies which are not specific to sequential data processing. Our results highlight the key role played by the sequence length and the importance of a clear specification of the CL scenario.

show abstract

Section: Survey Of Continual Learning In Recurrent Modelsmentioning

confidence: 99%

Continual Learning for Recurrent Neural Networks: an Empirical Evaluation

Cossu,

Carta,

Lomonaco

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…It is defined as the part of the prefrontal cortex that receives projections from the medial dorsal nucleus of the thalamus, and is thought to represent emotion and reward in decision making [28]. The prefrontal cortex, consisting of Brodmann areas 8,9,10,11,12,13,44,45,46 and 47, includes the OFC but covers a wider range of functionality.…”

Section: Reciprocitymentioning

confidence: 99%

“…The right side of Figure 4 (B) highlights reciprocal connections between cortical areas -Brodmann areas 7,8,9,11,12,13,19,20,21,22,23 and 46 -and the hippocampal complex via the adjacent perirhinal cortex (shown as blue connections) and the parahippocampal cortex (shown as red connections) that are involved in representing and recognizing objects and environmental scenes.…”

Section: Reciprocitymentioning

confidence: 99%

“…It may be possible to segment activity streams into coherent tasks in a similar way to how we segment conversations involving multiple speakers [165,166]. Alternatively, there has been some success with the method of pseudo rehearsal which consists of retraining existing networks by interleaving new examples with synthetic-examples produced by randomly activating the existing network [37,100,8,65,66,151].…”

Section: Action Selectionmentioning

confidence: 99%

See 1 more Smart Citation

Biological Blueprints for Next Generation AI Systems

Dean,

Fan,

Lewis

et al. 2019

Preprint

View full text Add to dashboard Cite

Diverse subfields of neuroscience have enriched artificial intelligence for many decades. With recent advances in machine learning and artificial neural networks, many neuroscientists are partnering with AI researchers and machine learning experts to analyze data and construct models. This paper attempts to demonstrate the value of such collaborations by providing examples of how insights derived from neuroscience research are helping to develop new machine learning algorithms and artificial neural network architectures. We survey the relevant neuroscience necessary to appreciate these insights and then describe how we can translate our current understanding of the relevant neurobiology into algorithmic techniques and architectural designs. Finally, we characterize some of the major challenges facing current AI technology and suggest avenues for overcoming these challenges that draw upon research in developmental and comparative cognitive neuroscience.

show abstract

“…The main problem with Creative is the rather large number of robots in L < 0. This might be attributed to Catastrophic Interference [10] in the memory, as many of these robots did not have any wall contact during pretraining. Forgetting of previous knowledge may also occur in the evaluation network, but with a second ANN (memory) in the neurocontroller this effect may be more pronounced.…”

Section: Resultsmentioning

confidence: 99%

A modular neurocontroller for creative mobile autonomous robots learning by temporal difference

Mayer

2004 IEEE International Conference on Systems, Man and Cybernetics (IEEE Cat. No.04CH37583)

View full text Add to dashboard Cite

One of the most prominent research goals in the field of mobile autonomous robots is to create robots that are able to adapt to new environments, i.e., the robots should be able to learn during their "lifetime" possibly without (or a minimum) of human intervention. When employing artificial neural networks (ANNs) to control the robot, reinforcement learning (RL) techniques are a good candidate for achieving continuous on-line learning. A problem with RL applied to robot learning is that the state (and action) space of a robot is typically not discrete. Thus, the robot had to evaluate an infinite number of possible actions at every time step in order to select the best. To overcome this problem we add a second network module to the neurocontroller acting as a memory of previous decisions (state-action pairs) of the robot. The robot's actual decisions, then, are based on previous decisions retrieved from memory. Additionally, intrinsic noise in the memory network gives the robot the possibility to evaluate new "ideas", hence it becomes creative. We analyze the potential of the above approach by measuring the ability of (simulated) robots to learn simple tasks using temporal difference (TD) learning.

show abstract

Preventing Catastrophic Interference in Multiple-Sequence Learning Using Coupled Reverberating Elman Networks

Cited by 9 publications

References 16 publications

Continual Learning for Recurrent Neural Networks: an Empirical Evaluation

Continual Learning for Recurrent Neural Networks: an Empirical Evaluation

Biological Blueprints for Next Generation AI Systems

A modular neurocontroller for creative mobile autonomous robots learning by temporal difference

Contact Info

Product

Resources

About