Welcome to Project Lovelace! We're still super new so there are still tons of bugs to find and improvements to make. If you have any suggestions, complaints, or any comments at all please let us know on Discourse!

You will learn about: Ada Lovelace, the first computer algorithm, and Bernoulli numbers.

Ada Lovelace (1815-1852) was an English mathematician who was the first person to realize that computers could be much more than big calculators and published the first complex computer algorithm in 1843. She did this long before any computer existed to execute her algorithm as it was devised for the Analytical Engine which never got built. The algorithm, detailed in Note G of Sketch of The Analytical Engine Invented by Charles Babbage by Luigi Menabrea with notes by Ada Lovelace described how the Analytical Engine could calculate the Bernoulli numbers using a recursive algorithm, which you will reproduce here.

The Bernoulli numbers $B_n$ are a sequence of rational numbers (fractions) that tend to show up in many places in mathematics, and there are many ways of calculating them. One method is to calculate them recursively where calculating $B_n$ involves knowing $B_0, B_1, \dots, B_{n-1}$. A derivation of how Ada Lovelace came up with the formula is posted in the notes tab, but she essentially used the following formula $$B_n = -\sum_{k=0}^{n-1} \frac{n!}{(n+1-k)!\cdot k!} B_k = -\sum_{k=0}^{n-1} \binom{n}{k} \frac{B_k}{n+1-k}$$ where $B_n$ is calculated from $B_0, B_1, \dots, B_{n-1}$. Here $\displaystyle \binom{n}{k} = \frac{n!}{k!(n-k)!}$ is a binomial coefficient, and $n! = n(n-1)(n-2)\cdots3\cdot2\cdot1$ is the factorial which is the product of all positive integers less than or equal to $n$. $B_0 = 1$ while there are two choices for $B_1$ corresponding to two conventions and we will choose $\displaystyle B_1 = -\frac{1}{2}$ which most modern textbooks use. Diagram of an algorithm for the Analytical Engine for the computation of Bernoulli numbers, from Sketch of The Analytical Engine Invented by Charles Babbage by Luigi Menabrea with notes by Ada Lovelace. The steps are describing how the analytical engine would compute $B_8$, which she called $B_7$ in her notation. (Source: Wikimedia Commons)

Write a function that calculates $B_n$ and returns the numerator and denominator. If $B_n$ is negative, the minus sign should be included with the numerator.

Input: An integer $n$.

Output: The numerator and denominator of the $n^\textrm{th}$ Bernoulli number $B_n$. If $B_n$ is negative, the minus sign should be included with the numerator. If $B_n = 0$ then return a numerator of 0 and a denominator of 1.

Example 1

Input n: 8 Output numerator: -1 Output denominator: 30

Example 2

Input n: 3 Output numerator: 0 Output denominator: 1

Example 3

Input n: 88 Output numerator: -1311426488674017507995511424019311843345750275572028644296919890574047 Output denominator: 61410

 Difficulty  Timesink 60 s 250 MiB bernoulli(n)

Write a function that accepts the input as function parameters and returns the correct output. Make sure to read the description above to produce the correct output in the correct format and use the correct function signature so we can run your code. A good first step is to try reproducing the example(s). Your code must not take longer than the maximum runtime to run and must not use more memory than the allowed limit.

You must be logged in to view your submissions.

### Derivation of Ada Lovelace's algorithm

To see how Ada derived her formula for the Bernoulli numbers we will have to make use of the fact that $e^t$ can be expressed as an infinite series $\displaystyle e^t = \sum_{n=0}^\infty \frac{t^n}{n!:}$ which is called its Taylor series. The Bernoulli numbers $B_n$ can be defined as the coefficients of the Taylor series of a generating function $$\frac{t}{e^t - 1} = \sum_{n=0}^\infty B_n \frac{t^n}{n!}$$ We now want to solve for $B_n$ to get a formula for the Bernoulli numbers we can use to program a computer. We move both terms to a single side and insert the Taylor series for $e^t$ to get $$1 = \frac{e^t - 1}{t} \sum_{n=0}^\infty B_n \frac{t^n}{n!} = \sum_{m=0}^\infty \frac{x^m}{(m+1)!} \sum_{n=0}^\infty B_n \frac{t^n}{n!}$$ We now have the product of two infinite series so let's expand their product up to terms including $t^4$. This gives us \begin{align*} 1 &= \left( 1 + \frac{t}{2} + \frac{t^2}{2\cdot3} + \frac{t^3}{2\cdot3\cdot4} + \frac{t^4}{2\cdot3\cdot4\cdot5} + \mathcal{O}(t^5) \right) \left( B_0 + B_1t + B_2 \frac{t^2}{2} + B_3\frac{t^3}{2\cdot3} + B_4\frac{t^4}{2\cdot3\cdot4} + \mathcal{O}(t^5) \right) \\ &= B_0 + \left( B_1 + \frac{B_0}{2} \right) t + \left( \frac{B_2}{2} + \frac{B_1}{2} + \frac{B_0}{2\cdot3} \right)t^2 + \left( \frac{B_3}{2\cdot3} + \frac{B_2}{2\cdot2} + \frac{B_1}{2\cdot3} + \frac{B_0}{2\cdot3\cdot4} \right) t^3 \\ & \quad + \left( \frac{B_4}{2\cdot3\cdot4} + \frac{B_3}{2\cdot2\cdot3} + \frac{B_2}{2\cdot2\cdot3} + \frac{B_1}{2\cdot3\cdot4} + \frac{B_0}{2\cdot3\cdot4\cdot5} \right) t^4 + \mathcal{O}(t^5) \end{align*} where we grouped terms with the same powers of $t$. Now we can start calculating the Bernoulli numbers. The left and right hand sides must be equal so $B_0 = 1$. As $t$ does not appear on the left, we can calculate $B_1$ using $$B_1 + \frac{B_0}{2} = 0 \quad \implies \quad B_1 = -\frac{B_0}{2} = -\frac{1}{2}$$ and similarly for $B_2$ as $t^2$ does not appear on the left $$\frac{B_2}{2} + \frac{B_1}{2} + \frac{B_0}{2\cdot3} \quad \implies \quad B_2 = -2\left( \frac{B_1}{2} + \frac{B_0}{6} \right) = -2\left( -\frac{1}{4} + \frac{1}{6} \right) = \frac{1}{6}$$ Repeating the process for the $t^3$ and $t^4$ terms we find $B_3 = 0$ and $\displaystyle B_4 = -\frac{1}{30}$. We also notice a pattern: the coefficient of $t^n$, which is equal to zero, is given by \begin{multline} 1\cdot\frac{B_n}{2\cdot3\cdot4\cdots(n+1)} + \frac{1}{2} \cdot \frac{B_{n-1}}{2\cdot3\cdot4\cdots n} + \frac{1}{2\cdot3} \cdot \frac{B_{n-2}}{2\cdot3\cdot4\cdots (n-1)} + \cdots \\ + \frac{1}{2\cdot3\cdots(n-2)} \cdot \frac{B_3}{2\cdot3} + \frac{1}{2\cdot3\cdots(n-1)} \cdot \frac{B_2}{2} + \frac{1}{2\cdot3\cdots n} \cdot B_1 + \frac{1}{2\cdot3\cdots(n+1)} \cdot B_0 = 0 \end{multline} which we can write as $$\sum_{k=0}^n \frac{1}{(n+1-k)!} \cdot \frac{B_n}{k!} = 0$$ and so if we take out the $B_n$ term we can express it in terms of $B_k$ for $k \lt n$ $$B_n = -\sum_{k=0}^{n-1} \frac{n!}{(n+1-k)!\cdot k!} B_k = -\sum_{k=0}^{n-1} \binom{n}{k} \frac{B_k}{n+1-k}$$ Using this formula along with $B_0 = 1$ we can calculate any Bernoulli number $B_n$ as long as we know all the lower Bernoulli numbers $B_0, B_1, B_2, \dots, B_{n-1}$. In Ada Lovelace's Note G, she proposes an algorithm for calculating what we call $B_8$, which she labeled $B_7$.

### Notes

Let us know what you think about this problem! Was it too hard? Difficult to understand? Also feel free to discuss the problem, ask questions, and post cool stuff on Discourse. You should be able see a discussion thread below. Feel free to post your solutions but if you do please organize and document your code well so others can learn from it.