Monkeyzx
Xiahouzuoxin Github. Theme PaperMod - https://github.com/adityatelange/hugo-PaperMod.
Read Xiahouzuoxin.github.io news digest here: view the latest Xiahouzuoxin Github articles and content updates right away or get to their most visited pages. Xiahouzuoxin.github.io is not yet rated by Alexa and its traffic estimate is unavailable. It seems that Xiahouzuoxin Github content is notably popular in China. We haven’t detected security issues or inappropriate content on Xiahouzuoxin.github.io and thus you can safely use it. Xiahouzuoxin.github.io is hosted with GitHub, INC. (Netherlands).
Content verdict: Safe
Website availability: Live
Last check:
-
N/A
Visitors daily -
N/A
Pageviews daily -
N/A
Google PR -
N/A
Alexa rank
Best pages on Xiahouzuoxin.github.io
-
Theme PaperMod - https://github.com/adityatelange/hugo-PaperMod
-
All Categories · zuoxin's notes
All categories of the zuoxin's notes blog
-
这些年我一直提醒自己一件事情,千万不要自己感动自己。大部分人看似的努力,不过是愚蠢导致的。什么熬夜看书到天亮,连续几天只睡几小时,多久没放假了,如果这些东西也值得夸耀,那么富士康流水线上任何一个人都比你努力多了。人难免天生有自怜的情绪,唯有时刻保持清醒,才能看清真正的价值在哪里。 ——于宙TEDx演讲:我们这一代人的困惑 平时总拿追求安逸平凡作为生活的目标,其实是给自己懒惰找的借口而已。为什么要读...
Xiahouzuoxin.github.io news digest
-
1 year
背景 想要用ML模型拟合业务,但实际中经常会遇到类似下面的场景:二手车价格预估:同一辆车,预估的价格与里程应该是严格负相关的; 保费预估:保额越高,保费越高; 理财产品中,比如余额宝,用户的利率敏感度预估(利率敏感度=用户余额变化/利率变化):同一用户,理财产品利率越高,购买金额越高,余额越高; 如果直接用比如xgb或dnn模型拟合 X -> Y,$$ p_y=f(x), where\ x = [x_c, x_p] $$ 其中f为xgb或dnn等具体模型,x_c为先验无关的特征(比如二手车里面的品牌和里程特征无关...
-
1 year
为什么需要重读word2vec Word2vec原论文:Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient estimation of word representations in vector space. CoRR, abs/1301.3781,2013. Tomas Mikolov, Ilya Sutskever, Kai Chen, Gregory S. Corrado, and Jeffrey Dean...
-
1 year
SIM和ETA的问题 SIM通过类目从长序列中检索相关的items(hard-search)或者通过embedding inner product计算从长序列中检索最相似的topK个item;SIM的问题在 ETA 中讲过,主要是离线索引可能带来的在线target item embedding和离线的item embeddings不一致的问题; ETA在SIM基础上,通过对Long Behavior Sequence的item embeddings进行SimHash(LSH)之后,然后就可以将inner...
-
1 year
SIM的问题 前面介绍过阿里巴巴超长行为序列建模的方法SIM是two-stage,在预估的时候,SIM先通过target item从长行为序列中通过类目(hard-search)或item embedding相似度(soft-search)检索出top-K,再送到主模型中做target attention。由于类目索引和item embedding索引的构建是离线的,而CTR主模型很多时候需要online learning,这种情况下就造成用于检索的目标item embedding和离线构建的item...
Domain history
| Web host: | GitHub, INC. |
| Registrar: | |
| Registrant: | REDACTED (GitHub, Inc.) |
| Updated: | February 09, 2025 |
| Expires: | March 08, 2027 |
| Created: | March 08, 2013 |
Whois record
Safety scores
Trustworthiness
N/AChild safety
N/A