RE: Math from Finland, yes, it's fast: FastICA by terrylovejoy

View this thread on steempeak.com

Viewing a response to: @alexs1320/math-from-finland-yes-it-s-fast-fastica

· @terrylovejoy ·
$0.06
Thank you for the shoutout @alexs1320 !  I had to go away and think about how this would be applied to image processi  ng (for astronomical images of course) and there could be some interesting applications to explore.  From what I read this is a technique of estimating the original contributions of multiple signals embedded in a single signal.  I will definitely keep this and your links bookmarked!
👍  
properties (23)
post_id47,351,538
authorterrylovejoy
permlinkre-alexs1320-math-from-finland-yes-it-s-fast-fastica-20180507t085604484z
categorysteemstem
json_metadata"{"app": "steemit/0.1", "users": ["alexs1320"], "tags": ["steemstem"]}"
created2018-05-07 08:56:06
last_update2018-05-07 08:56:06
depth1
children1
net_rshares12,289,634,901
last_payout2018-05-14 08:56:06
cashout_time1969-12-31 23:59:59
total_payout_value0.041 SBD
curator_payout_value0.014 SBD
pending_payout_value0.000 SBD
promoted0.000 SBD
body_length402
author_reputation14,677,992,676,220
root_title"Math from Finland, yes, it's fast: FastICA"
beneficiaries[]
max_accepted_payout1,000,000.000 SBD
percent_steem_dollars10,000
author_curate_reward""
vote details (1)
@alexs1320 ·
(Probably) the best algorithm would be the SOBI-RO from the package ICALAB, by Andrzej Cichocki. That obscure method works with everything I've ever put as the input. It's especially good for sparse signals.

In your case, the workflow would be: each image --> image to vector (in Matlab) --> all the vectors packed into the input matrix --> SOBI-RO --> back to images (vector to matrix) --> and... see what is the signal and what is the noise :) --> discard the noise 

From *n* images, *n* components could be extracted (both in ICA and SOBI), in contrast to PCA/FA where you need *2n+1* starting signals for only *n* components,
properties (22)
post_id47,357,136
authoralexs1320
permlinkre-terrylovejoy-re-alexs1320-math-from-finland-yes-it-s-fast-fastica-20180507t094617112z
categorysteemstem
json_metadata"{"app": "steemit/0.1", "tags": ["steemstem"]}"
created2018-05-07 09:46:21
last_update2018-05-07 09:46:21
depth2
children0
net_rshares0
last_payout2018-05-14 09:46:21
cashout_time1969-12-31 23:59:59
total_payout_value0.000 SBD
curator_payout_value0.000 SBD
pending_payout_value0.000 SBD
promoted0.000 SBD
body_length631
author_reputation149,050,464,338,619
root_title"Math from Finland, yes, it's fast: FastICA"
beneficiaries[]
max_accepted_payout1,000,000.000 SBD
percent_steem_dollars10,000