Beumer, Lhea (2025) Credit Assignment in Cortex: Backpropagation or Target Learning? Master's Thesis / Essay, Computational Cognitive Science.
|
Text
mCCS2025LheaBeumer.pdf Download (4MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming beumer.pdf Restricted to Registered users only Download (193kB) |
Abstract
Understanding how networks of cortical neurons learn by adjusting synaptic weights remains one of the greatest frontiers of neuroscience. Machine learning theory has proposed two opposing hypotheses. Backpropagation-based (BP) algorithms that rely on minimizing error signals, and Target-based Learning (TL) algorithms in which target activity states are consolidated. This thesis provides a novel framework for experimentally distinguishing between these two theories of cortical learning and moves beyond purely theoretical arguments to concrete, data-driven validation of hierarchical learning in mouse lateral visual cortex. By showing that the reactivations of stimulus‑evoked responses serve as a teaching signal and causally shape future responses, we are able to differentiate inference from learning phases. Furthermore, we reveal that the neural dynamics underlying these two phases is qualitatively distinct, which contradicts a key assumption of neuro‐plausible BP approximations. Finally, results indicate that reactivation-driven plasticity is invariant to stimulus identity, leading to the hypothesis that reactivations might communicate a general teaching signal, rather than a stimulus-specific label. Collectively, we argue that these findings indicate that cortical hierarchical learning operates in a TL-compatible and BP-incompatible regime. Through the combination of causal, dynamical and representational analyses, this work furthers understanding of cortical hierarchical com
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Vugt, M.K. van |
| Degree programme: | Computational Cognitive Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 11 Jul 2025 14:11 |
| Last Modified: | 11 Jul 2025 14:11 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/35898 |
Actions (login required)
 |
View Item |