Frederik Petersen
6c9187ec95
Added new regression type
2015-02-09 11:12:37 +01:00
Frederik Petersen
0a10483110
Merge branch 'master' into polars-multivariat-lin-reg
2015-02-09 08:59:53 +01:00
Simon Pamies
7288aca296
Back to the name EXPEDITION for measured wind sources - a change here has too many implications such as needing to migrate the database
2015-02-08 15:29:08 +08:00
Simon Pamies
435a562c2e
Updated wind source API to also deliver wind source ids to make it possible to filter them; also renamed expedition to windSource
2015-02-08 14:45:53 +08:00
Axel Uhl
04775d34f0
fixed bad bug in finding head-up/bear-away clusters (swapped max for min); consider head-up/bear-away clusters for jibe cluster likelihood
2015-01-27 17:24:50 +01:00
Axel Uhl
369a48f44a
preparing for jibe head-up/bear-away cluster detection; fixing test verdict for 2011 KW 505 R4 wind direction
2015-01-27 13:33:04 +01:00
Axel Uhl
86fb24201b
factored finding of head-up/bear-away clusters into separate method
2015-01-26 18:59:59 +01:00
Axel Uhl
bef616d474
considering head-up / bear-away clusters on port and starboard tack on the upwind legs for tack cluster likelihoods
2015-01-26 18:14:57 +01:00
Axel Uhl
500dcf41e9
cache ScalableBearingAndScalableDouble for ManeuverClassification, avoiding expensive creation; use approximation for ScalableBearing difference which is much faster
2015-01-24 02:24:59 +01:00
Axel Uhl
882f18300f
fixed caching for likelihoodAndTWSBasedOnSpeedAndAngle; it had to be maneuver type-specific
2015-01-24 01:37:10 +01:00
Axel Uhl
13a6ea0ced
introduced caching for likelihoodAndTWSBasedOnSpeedAndAngleCache on ManeuverClassification
2015-01-24 01:32:52 +01:00
Axel Uhl
f3499f6d6b
improved tack cluster likelihood by looking for the opposite tack cluster; OBMR test for small fleet green now
2015-01-24 01:07:00 +01:00
Axel Uhl
a3af0faae3
measure time it takes to compute wind based on maneuvers; re-enabled two more tests of this kind
2015-01-23 17:46:33 +01:00
Axel Uhl
eb8c8b2537
changed from one-dimensional clustering along middle COG to two-dimensional clustering (middle COG / maneuver angle) and starting first tests with this approach
2015-01-23 17:29:54 +01:00
Axel Uhl
abe1092dae
Merge branch 'polar-api' into bug1562
2015-01-22 23:40:41 +01:00
Axel Uhl
db68200280
moved maneuver angle likelihood judgement from wind estimation feature to PolarDataService
2015-01-22 23:34:36 +01:00
Frederik Petersen
7801298269
Some changes to the online regression
...
input normalization is not functioning yet
2015-01-22 19:11:40 +01:00
Axel Uhl
70ad5b33de
added tack and jibe likelihood to debug output
2015-01-22 17:58:17 +01:00
Axel Uhl
95f6264ed4
a first two-dimensional clustering algorithm
2015-01-22 14:01:16 +01:00
Axel Uhl
6c2d8238cf
avoid warning by declaring Util.getFirstNonNull with @SafeVarargs; this was necessary after the upgrade from using JDK 1.6 --> 1.7
2015-01-22 11:29:50 +01:00
Frederik Petersen
393a5e1b37
Adjusted testing scenario and regression itself to christophers R example
2015-01-22 11:04:23 +01:00
Axel Uhl
9fb859a8b3
major refactoring, solving bug 2578 by introducing com.sap.sse.shared.android and moving useful stuff to it that will compile on Android
2015-01-22 10:57:00 +01:00
Axel Uhl
76e52fefab
added comments
2015-01-22 09:00:15 +01:00
Axel Uhl
dd60863858
moved scalablevalue stuff and KMeansClusterer to SSE; added better maneuver-based wind estimation reporting features
2015-01-22 00:49:34 +01:00
Axel Uhl
14412c8290
got the first working test case for maneuver-based wind inference; two more tests still fail with a slightly mismatching wind estimation; need to analyze further
2015-01-21 21:29:04 +01:00
Frederik Petersen
be98b84324
Adjusted regression algorithm
...
-used constant alpha
-different test case
-still a problem with huuuge weigths
2015-01-21 18:19:18 +01:00
Frederik Petersen
7504986f6b
Added normalization of input for regression
...
also added a test. it fails at this point due to some error..
2015-01-21 13:17:38 +01:00
Frederik Petersen
4e207c872d
Added implementation of online multi-variate linear Regression
2015-01-21 11:05:19 +01:00
Axel Uhl
0af5b465b7
slight modification of simplified getLikelihoodTackJibeSpeedRatio formula
2015-01-21 10:58:35 +01:00
Frederik Petersen
ba06f8f577
Added interface for multivariate online regression
2015-01-21 10:36:58 +01:00
Axel Uhl
393d2d1246
do a first basic speed comparison for the tack and jibe cluster; multiply instead of add likelihoods of tack / jibe cluster (see bug 1562 comment #6 )
2015-01-21 01:05:28 +01:00
Axel Uhl
f5e48920b8
made a few adjustments mostly regarding null-safety
2015-01-21 00:20:23 +01:00
Axel Uhl
0ccd2629e8
implemented a first outline of the algorithm described in a comment in an earlier commit
2015-01-21 00:02:04 +01:00
Axel Uhl
8264e1d3c4
added TODO comments
2015-01-20 18:20:28 +01:00
Axel Uhl
1a62ca12c9
added comments to ManeuverBasedWindEstimationTrackImpl for intended implementation
2015-01-20 18:16:37 +01:00
Axel Uhl
a335401592
modified PolarResource such that it output maneuver data ready for Symbiosis
2015-01-20 11:24:44 +01:00
Axel Uhl
2d1a4b6b6a
commented on some new implementation ideas
2015-01-14 18:36:24 +01:00
Axel Uhl
6e75407d02
fixed testMappingClusterer(), establishing a stable set of seeds
2015-01-14 17:51:10 +01:00
Axel Uhl
7bb96f1bd2
enhancing documentation in ManeuverBaseWindEstimationTrackImpl, preparing for better confidence handling
2015-01-14 15:47:34 +01:00
Axel Uhl
89e83a6955
Merge branch 'master' into bug1562
2015-01-14 14:56:30 +01:00
Frederik Petersen
a1e73cee71
getAverageTrueWindSpeedAndAngleCandidates stub changed
...
if a windspeed candidate is over-or-eq to 40kn it will not be added to
result set. the cubic functions of the stub are unrealistic beyond the
40kn windspeed mark
2015-01-14 10:45:18 +01:00
Axel Uhl
29d58e03d6
worked further towards speed-based post-processing of flip cluster
2015-01-14 01:27:03 +01:00
Axel Uhl
651cb52e58
working on clustering and classifying the different maneuvers
2015-01-14 00:41:49 +01:00
Axel Uhl
a06c18e31a
use KMeansMappingClusterer to cluster maneuver classifications by their maneuver angle, finding dominant and opposite angle
2015-01-13 18:24:54 +01:00
Axel Uhl
d3130c9401
introduced a k-means clusterer that uses a mapper function; not entirely finished yet
2015-01-13 17:52:51 +01:00
Axel Uhl
be843c8ddf
working on the KMeansClusterer
2015-01-13 12:44:46 +01:00
Axel Uhl
320e1bcfaf
started to implement a K-Means clustering algorithm
2015-01-13 11:25:48 +01:00
Axel Uhl
93afebae8f
record and show maneuver classifications per leg type / tack and add competitor and time point info
2015-01-12 09:46:33 +01:00
Axel Uhl
8e4113a8f2
improved logging output for maneuver-based wind estimation
2015-01-08 16:28:52 +01:00
Axel Uhl
c86ce4ab01
avoid a few NPEs in polar sheet calculation
2015-01-07 21:03:17 +01:00
Frederik Petersen
Simon Pamies
Axel Uhl