Which statement is true according to the material?

• A. They cannot model complex relationships
• B. They are easier to interpret than decision trees
• C. They do not require data to train
• D. Neural networks excel at finding the strength of connections between attributes

Answer: (D)

Neural networks learn by adjusting weights that measure how strongly one feature’s signal influences the next unit. During training, these weights are tuned to minimize error, and with multiple layers they combine features in nonlinear ways to capture complex relationships. This makes the statement about finding the strength of connections between attributes the best description: the weights explicitly encode how strongly each input affects downstream computations, across the network.

The other ideas don’t fit as well: neural networks are designed to model nonlinear, intricate patterns; they’re not typically easier to interpret than a decision tree; training requires data to learn from; and the key takeaway is about learning connection strengths, not about existing without data or being inherently easy to interpret.