SophieZhou的专栏

<br />概率图模型是一个迅速在国内得到关注的东东。我想可能是因为我们这做机器学习的一个老师最后组内做了个总结，现在好多单位都开始了。清华都已经开始上课了，真是够迅速的啊，本来觉得自己开始的很早呢，可现在看来只是比别人貌似早了大概一年不到，再加上自己水平低，开始的起点低，看来再不跑就被人追上了。

<br />Bayesian models are flexible to prior.<br /> <br /> <br />I am a little sad about the Bayesian prior. The experiments are so sensitive to the prior. this is not my expectation. <br /> <br />In my opinion, I have not considered this before, and I thin