machine learning - Conclusion from PCA of dataset -


i have set of data sequence labeling. did pca (with 2 principal components on x , y axis) on dataset , turns out below: pca of raw dataset

using lstm network classify dataset above, decided extract activations hidden layer of lstm. obtain figure below: pca of activation lstm hidden layer using above dataset

my question is, conclusion can draw comparing both results? fair features of original dataset self-organized after running through lstm classifier?


Comments

Post a Comment

Popular posts from this blog

java - BeanIO write annotated class to fixedlength -

i trying write annotated class fixedlength file beanio. classes annotated getting exception "invalid field 'employees', in record 'team', in stream 'tm': type handler not found type 'com.mycompany.bio.employee'" below source code public static void main(string[] args) { streamfactory factory = streamfactory.newinstance(); streambuilder builder = new streambuilder("tm") .format("fixedlength") .parser(new fixedlengthparserbuilder()) .addrecord(com.mycompany.bio.team.class); factory.define(builder); employee e1 = new employee("empf1", "empl1", "developer", "1", new date()); employee e2 = new employee("empf2", "empl2", "developer", "2", new date()); team team = new team(); team.setteamname("great team"); team.getemployees().add(e1); team.getemployees().a...
Read more

Using Java 8 lambdas/transformations to combine and flatten two Maps -

i have 2 maps: map<a, collection<b>> mapab map<b, collection<c>> mapbc i transform them map<a, collection<c>> mapac , i'm wondering if there's smooth way lambdas , transformations. in particular case, collections sets, i'd solve problem collections in general. one thought had first combine 2 maps map<a, map<b, collection<c>>> , flatten it, i'm open approach. data notes: b should occur in value collection associated 1 a , , same true mapbc (a given c mapped 1 b ). result, there should 1 path given a given c , although there may a -> b mappings there no b -> c mappings , there may b -> c mappings there no corresponding a -> b mappings. these orphans don't appear in resulting mapac . for sake of comparison, here's example of purely imperative approach same problem: map<a, collection<c>> mapac = new hashmap<>(); (entry<a, collection<b>> entry...
Read more

How to connect android app to App engine -

i'm trying connect app in android server i've opened on appengine. my server's name : dddd-daniel-2345 , website server : http://dddd-daniel-2345.appspot.com/ this how connect android app server in app engine: app.yaml: application: dddd-daniel-2345 version: 1 runtime: python27 api_version: 1 threadsafe: yes handlers: - url: .* script: main.app libraries: - name: webapp2 version: "2.5.2" main.py: #!/usr/bin/env python # # copyright 2007 google inc. google.appengine.api import users import webapp2 class userhandler(webapp2.requesthandler): def get(self): user = users.get_current_user() if user: greeting = ('%s,%s,%s' % (user.email(),user.user_id(),user.nickname())) else: greeting = ('not logged in') self.response.out.write(greeting) app = webapp2.wsgiapplication([('/', userhandler),], debug=true) my applications code : add manifast main....
Read more