Closed murdonson closed 7 years ago
Hi, I checked the reference paper and found that the original implementation is the same with the paper, however your version loses "(trustedValue / numTrusted)" for "userFactors.get(trustedUserIdx, factorIdx)".
As shown in the picture below :
嗯嗯 我搞错了 3q
我个人觉得应该是这样啊 -regSocial (userFactors.get(trustedUserIdx, factorIdx) -(trustedValue / numTrusted) sumDiffs[factorIdx] / numTrust) 按照 我那样改后 指标有一点点的提升 对不 亚童?