Hi, Adapters:<div><p class="MsoNormal"><span style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">  <span lang="EN-US">I will introduce a manifold
learning method t-SNE, which can be regarded as dimensionality  reduction
,I think.<o:p></o:p></span></span></p>

<p class="MsoNormal"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">  There exist many methods about
 dimensionality  reduction, some linear like Principal Component
Analysis(PCA), Linear Discriminant Analysis(LDA)бн, some nonlinear like Isomap, LoLaplaciancally
Linear Embedding(LLE)бн. and I will introduce one nonlinear method t-SNE, by
which The visualizations produced are significantly better than those produced
by the other techniques on almost all of the datasets.<o:p></o:p></span></p>

<p class="MsoNormal"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">   The idea of this algorithm is clear, But there
still exist many details deserved to be understood. I hope you can have some
review of probability theory before, such as Gaussian distribution, conditional
probability, and t-distribution. And KL divergence, which is also called
relative entropy, is also needed.</span></p>

<p class="MsoNormal" style="text-indent:5.25pt;mso-char-indent-count:.5"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;"> Here
 are some useful references ordered by
importance:<o:p></o:p></span></p>

<p class="MsoNormal" style="text-indent:21.0pt"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">Maaten L, Hinton G.
Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008,
9(Nov): 2579-2605.</span></p><p class="MsoNormal" style="text-indent:21.0pt"><span lang="EN-US" style="background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial; font-family: Verdana, sans-serif;">http://bindog.github.io/blog/2016/06/04/from-sne-to-tsne-to-largevis</span></p><p class="MsoNormal" style="text-indent:21.0pt"><span lang="EN-US" style="background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial; font-family: Verdana, sans-serif;">http://blog.pluskid.org/?p=533</span></p><p class="MsoNormal" style="text-indent:21.0pt"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">Best Wishes</span></p><p class="MsoNormal" style="text-indent:21.0pt"><span lang="EN-US" style="font-family: Verdana, sans-serif; background-image: initial; background-position: initial; background-size: initial; background-repeat: initial; background-attachment: initial; background-origin: initial; background-clip: initial;">Haijun</span></p></div>