Sign in Subscribe
Artificial intelligence

Artificial Neural Networks Aligned with Brain Dynamics Show Promise in Enhanced Performance

Strategic Insights

— 2 min read

Researchers at Centre de Recherche de l'Institut Universitaire de Geriatrie de Montreal have made significant strides in developing artificial neural networks (ANNs) that are aligned with brain dynamics. The study involved training ANNs in the field of artificial intelligence (AI) to model brain processes, with the goal of enhancing performance on AI tasks. By aligning network representations with brain dynamics, the researchers achieved substantial improvements in brain encoding and downstream task performance.

Key Takeaways:

  • The study aimed to answer two key questions: Can brain alignment of auditory models lead to improved brain encoding for novel, previously unseen stimuli? Can brain alignment lead to generalisable representations of auditory signals that are useful for solving a variety of complex auditory tasks?
  • The researchers used two massive datasets: a deep phenotyping dataset from the Courtois neuronal modelling project and the HEAR benchmark, a large battery of downstream auditory tasks.
  • Fine-tuning a small pretrained convolutional neural network with ~2.5 M parameters and aligning it with brain data from three seasons led to substantial improvement in brain encoding in the fourth season, extending beyond auditory and visual cortices.
  • The study observed consistent performance gains on the HEAR benchmark, particularly for tasks with limited training data, where brain-aligned models performed comparably with the best-performing models regardless of size.
  • The research highlighted the data efficiency of fine-tuning with individual brain data, with individual models often matching or outperforming group models in both brain encoding and downstream task performance.

Statistics:

  • The research used a deep phenotyping dataset from the Courtois neuronal modelling project, where six subjects watched four seasons (36 h) of the TV series in functional magnetic resonance imaging.
  • The study fine-tuned a small pretrained convolutional neural network with ~2.5 M parameters.
  • The research observed substantial improvements in brain encoding in the fourth season, extending beyond auditory and visual cortices.
  • The study saw consistent performance gains on the HEAR benchmark, particularly for tasks with limited training data, where brain-aligned models performed comparably with the best-performing models regardless of size.
  • The research highlighted the data efficiency of fine-tuning with individual brain data, with individual models often matching or outperforming group models in both brain encoding and downstream task performance.

Sources:

  • "Alignment of auditory artificial networks with massive individual fMRI brain data leads to generalisable improvements in brain encoding and downstream tasks. Imaging Neuroscience, 2025;3."
  • NewsRx. Study Data from Centre de Recherche de l'Institut Universitaire de Geriatrie de Montreal Provide New Insights into Artificial Neural Networks (Alignment of auditory artificial networks with massive individual fMRI brain data leads to ...). Journal of Engineering. August 25, 2025; p 3814.

Read more