Search the Community
Showing results for tags 'neutral networks'.
Found 1 result
As Valve's Steam Labs launch three new experimental features today, one has caught the interest of many Steam users: its new algorithm for game recommendations based on Valve's machine learning technology. Valve says the Interactive Recommender uses a "neural-network model that is trained to recommend games based on a user's playtime history, along with other salient data." The data is modified by two sliders that users can edit: one ranges from "popular" to "niche," while the other slider ranges from "older" to "newer" games. Rather than base recommendations around genre or category, the Interactive Recommender instead scans through Valve's data sets to find other Steam users with similar tastes. The model then recommends titles the user might enjoy based on other games played by like-minded Steam users. Valve also says they discard most category information about the game when entering it into their model. "We don't explicitly feed our model information about the games. Instead, the model learns about the games for itself during the training process. In fact, the only information about a game that gets explicitly fed into the process is the release date, enabling us to do time-windowing for the release-date slider. It turns out that using release date as part of the model training process yields better quality results than simply applying it as filter on the output," Valve said. They also discard information about review scores and tags, relying only on popularity and age variables. Users worried about this experimental technology replacing their regular Steam recommendations have nothing to fear for the time being. Rather, Valve says users who want to try the Recommender will have to specifically choose it under the Steam Labs experiments section. Regular Steam recommendations will still function as before. Since their algorithm discards the categories most other game recommendation algorithms operate by, Valve also claims that developers won't have to worry about optimizing their game description to make it more likely to be recommended. "The best way for a developer to optimize for this model is to make a game that people enjoy playing. While it's important to supply users with useful information about your game on its store page, you shouldn't agonize about whether tags or other metadata will affect how a recommendations model sees your game," Valve said. If you want to try the Interactive Recommender, head over to the Steams Labs experiments section. (via PC Gamer)