evolutioner

3 ms·

1. 1.▲
   by evolutioner 9y ago

   You're right, the paper linked above is a different way of solving the same problem. In their case they use a model to decide which website variants to show. Their model accounts for independent effects and pairwise dependencies. Evolu
2. 2.▲
   by evolutioner 9y ago

   They can actually work together. Thompson Sampling / Multi-armed bandit algorithms are traffic allocation strategies to use the least amount of traffic to find the optimal variant. So say you have version A/B/../N varian
3. 3.▲
   by evolutioner 9y ago

   Each generation has like 10~20 candidate websites. The better the conversion rate, the more likely it is to get chosen as a parent. With your example, let's say two parents with top/blue and bottom/red are chosen, then their
4. 4.▲
   by evolutioner 9y ago

   Sentient employee here. I'll give an example on a problem for which we use evolutionary algorithms: website optimization. Say you want to try many various changes like the title of your page, the color of the background, the position o