A note on the capacity of the binary perceptron. (arXiv:2401.15092v1 [math.PR])
![A note on the capacity of the binary perceptron. (arXiv:2401.15092v1 [math.PR])](https://cdn.hashnode.com/res/hashnode/image/upload/v1704026789016/QS9k8VMZb.jpg)
Determining the capacity $\alpha_c$ of the Binary Perceptron is a long-standing problem. Krauth and Mezard (1989) conjectured an explicit value of $\alpha_c$, approximately equal to .833, and a rigorous lower bound matching this prediction was recently established by Ding and Sun (2019). Regarding the upper bound, Kim and Roche (1998) and Talagrand (1999) independently showed that $\alpha_c$ < .996, while Krauth and Mezard outlined an argument which can be used to show that $\alpha_c$ < .847. The purpose of this expository note is to record a complete proof of the bound $\alpha_c$ < .847. The proof is a conditional first
