Money Matters

The Best Movies On Netflix: Fruit Flies Hold The Key To A Better Recommendation

(Source: www.forbes.com)

Netflix

Researchers have found what could be the next breakthrough in search engines and algorithms like those found on Netflix when you search for the best movies to watch next.

Whether you recognize it or not, practically every online service you use employs an algorithm to guess and suggest alternate interests. From Spotify recommending a new playlist to Amazon suggesting you buy ski goggles after purchasing skis, businesses rely on their ability to draw you into peripheral interests.

To do this, most algorithms rely on their ability to match different products, queries, or services with one another, known as similarity searches. The degree to which an algorithm can do this accurately and quickly provides it an upper hand compared to competitors.

It turns out that fruit flies do an approximation of this, built into their DNA as a survival mechanism. When a fruit fly initially smells an odor the fruit fly categorizes that odor into something it should approach or avoid. They do this by similarity searches in comparing the odor with previously experienced odors and whether they were food or not.

It winds up being the simple fruit fly is much more efficient and accurate at similarity searches as compared with most similarity algorithms companies employ.

The way most similarity algorithms work is to categorize a product or query by assigning it “hashes” which identify the key characteristics of the product. These hashes are used to describe the product and when conducting a similarity search, the algorithm looks for similar hashes from the source product.

While conducting the locality-sensitive hashing for similarity searches, however, the hashes are often simplified down 10 or so for simplicity sake. While this improves speed processing, it limits the number of uniquely identifiable features of each product.

Fruit flies, on the other hand, after smelling a new odor fire 50 neurons in a unique combination (hashes). The fruit fly then expands this “description” of the odor out to 2,000 neurons, allowing them to much more accurately describe the new odor. The fruit fly then chooses the top 5 percent most descriptive neurons or hashes and stores those into longterm memory.

The alternate approach to a similarity search, expanding descriptors compared to simplifying descriptors, has been experimentally tested in algorithms. The research team used the fruit fly method in three test search algorithms. They found that the fruit fly approach improved performance and matches.

Lead author of the new study, Saket Navlakha, describes it as “say you have a bunch of people clustered by their relationships, and they’re bunched into a crowded room. Then take the same people and relationships, but have them spread out on a football field.”

“It will be much easier to see the structure of relationships and draw boundaries between groups in the expanded space relative to the crowded space.”

By expanding the neurons they ensure that the fruit fly is better able to distinctly identify different groups. As search and similarity algorithms continue to advance in their ability to guess what we want to buy or do next, they find an unconventional ally, the tiny fruit fly.

More Info: www.forbes.com

Advertisements