Implement the sigmoid function by hand using the formula 1 / (1 + exp(-x)), rather than a built-in activation call.
1 / (1 + exp(-x))
solve(...)