Towards generalization in subitizing with neuro-symbolic loss using holographic reduced representations

Key Points

• The study demonstrates enhanced generalization in subitizing through neuro-symbolic loss methods, improving accuracy.
• Significant improvements include a 30% increase in correct number recognition rates in cognitive models.
• Observational analysis explores holographic representations and their impact on subitizing tasks across various conditions.
• The findings suggest potential advancements in cognitive modeling but require further exploration in diverse settings.