Below are two software puzzles we'd like you to solve and develop against. Please read through each description and provide a solution that best satisfies the requested requirements.
Write some code, in a language of your choice, that will correctly solve arbitrary serial division problems such as the following:
[ [ 16 ÷ [ 8 ÷ 2 ] ÷ 4 ] ÷ 2 ÷ 80 ]
For simplicity the equations should be represented as nested arrays of doubles. So the previous example would be represented as:
[[16,[8,2],4],2,80]
Be sure to create a driver function that takes in a parameter of nested arrays of doubles.
You’ve recently come on board at our budding start up! In an effort to get to know each team member you decide to take each teammate out to lunch over the course of your first year. Everyone in the office considers themselves big foodies and have kept a record of their likes and dislikes towards restaurants in the area. To woo each teammate you decide to enact some crafty engineering to find an optimal restaurant recommendation that he or she has not been to for each lunch. The algorithm you select is a variant of collaborative filtering using a Jaccard similarity index.
Using provided JSON files seed/out/ratings.json, seed/out/teammates.json and seed/out/restaurants.json create a program that takes in all files and a teammate id as input to display an output of the top 3 restaurants to recommend to that teammate, ordered descending by their prediction score. To find these recommendations you will need to use the following two equations.
- Finding a similarity index between Teammate 1 and Teammate 2 with bounds between -1.0 and 1.0
The equation returns a similarity index between two teammates by first finding the intersection between all the likes for teammate 1 and teammate 2. Next, add the intersection of all the dislikes for teammate 1 and teammate 2. Then subtract the previous sum by the intersection of the likes of teammate 1 between the dislikes of teammate 2. Then subtract the result by the intersection of the dislikes of teammate 1 between the likes of teammate 2. Finally, divide the previous total by the union of all the ratings of teammate 1 between all the ratings of teammate 2.
- Prediction between each teammate and a specific restaurant
This equation returns a prediction score between a teammate and specific restaurant. First, sum up the similarity indexes between a specific teammate and all other teammates that have liked the restaurant. Next, take the difference of that sum by the sum of similarity indexes between the specific teammate and all other teammates that have disliked the restaurant. Finally, divide by the total number of teammates that have liked or disliked the restaurant.
- The similarity index has bounds between -1 and 1
- You can’t recommend a restaurant to a teammate they’ve already been to
- Correctness
- Readability of your code
- Test coverage
- Performance (but don't get too hung up here, after all "premature optimization is the root of all evil")
- Clean, commented and maintainable code
- Docs
You may write each solution in whatever language you are most comfortable with. Your solution must include usage instructions, this includes how to run and/or build.
Place your division solver application under a solver/ directory within this directory. Add your solution to the recommendation application under a recommendation/ directory within this directory. Then add to either a public github repo and share with us the url OR zip up this directory and email the submission back to us.