On the stability of learning
Why sparse feedback breaks learning even when the math says it should work
Learning becomes unstable when feedback signals are sparse and noisy—even if the algorithm theoretically learns the right answer on average. Researchers found that variance in learning can grow without bound in these conditions, showing that unbiased learning signals alone aren't enough; they also need to arrive frequently and clearly enough.
This explains a real constraint on how biological brains can learn. Neurons receive feedback only occasionally and imperfectly, not constantly, which means evolution must have built in mechanisms to keep learning stable under these conditions. Understanding these limits helps neuroscientists explain why biological learning rules look different from the algorithms we use in AI, and could guide better models of how brains actually work.