Multitasking Capability Versus Learning Efficiency in Neural Network Architectures

By Sebastian Musslick, Andrew M. Saxe, Kayhan Özcimder, Biswadip Dey, Gregory Henselman, J. Cohen. 2017

46 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Orthogonal representations for robust context-dependent task performance in brains and neural networks (2022, positive): “One possibility is that, over prolonged training, the dimensionality of neural representations is tailored to the transfer demands of the paradigm (Musslick et al., 2017).”
Efficiency of learning vs. processing: Towards a normative theory of multitasking (2020, background): “Rather than using a logistic function to characterize learning, it may be more reasonable to scale the beneﬁt of shared representations by the number of tasks (e.g. as in Musslick et al., 2017), or to implement the learning dynamics of actual neural networks on similar task spaces.”
Direct-fit to nature: an evolutionary perspective on biological (and artificial) neural networks (2019, background): “Such constraints may include the need to integrate information across modalities and timescales, while selecting and executing a small set of coherent behaviors at each moment (Musslick et al., 2017).”
The dimensionality of neural representations for control (2020, unknown): “For example, recent work using neural network models has demonstrated a basic tradeoff between generalizability and interference in multitasking [30 ,31,32].”
If deep learning is the answer, what is the question? (2020, background): “One fascinating theoretical perspective argues that the capacity limits that are inherent in biological control processes are a response to this conundrum 136 .”
Multiple task-demands flexibly optimize neural geometry in human ventral temporal cortex (2026, unknown): “To achieve an optimal representational geometry for the task at hand, while minimizing interference with other tasks, the brain could reconfigure representations based on task demands 8,9 .”