Demonstrations of neural network computations involving students

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• David Marr's three levels of analysis (implementational, algorithmic, computational) are crucial for understanding information processing systems.
• Neuroscience education commonly covers implementational (neurophysiology) and computational (systems/cognitive neuroscience) levels.
• The algorithmic level, explaining how low-level mechanisms yield high-level cognition, is often neglected, creating an explanatory gap.

Purpose of the Study:

• To address the neglect of the algorithmic level in neuroscience education.
• To demonstrate how neural networks can bridge the implementational and computational levels of analysis.
• To provide resources for instructors to teach information processing in neurons.

Main Methods:

• Utilizing neural network models to simulate information processing.
• Developing demonstrations that illustrate the link between neural activity and cognitive phenomena.
• Focusing on the flow of ions (e.g., sodium) as an example of implementational detail leading to cognitive function.

Main Results:

• Neural networks effectively demonstrate how collections of neuron-like units can produce cognitive phenomena.
• The proposed demonstrations facilitate understanding of the algorithmic level of analysis.
• Bridging the gap between neurophysiology and cognitive neuroscience is achieved through these models.

Conclusions:

• Neural networks are valuable tools for teaching the algorithmic level of brain function.
• Integrating these demonstrations can enhance student comprehension of how neural processes support cognition.
• Addressing the explanatory gap improves the completeness of neuroscience education.