Comments (13)
Ok, this is pretty embarrassing!
Indeed... I think a very safe default would be to just consider an extra class "unknown" with one line containing {}
as features:
unknown.txt
{}
Since class balancing is used, it should work really well. I added the "unknown.txt" to my data, and it seems to work.
I guess it could simply be added in a flag for training; whether it should create the unknown.txt file, or, otherwise delete if existing.
from whereami.
If I'm not mistaken, unknown values when working with properly learned data is just when the confidence of every location is low.
That is to say, I think "unknown location" is equivalent to "the confidence for all other locations is below a threshold" what that threshold would be I have no idea.
from whereami.
@kionay Usually, the probabilities across classes would add up to 1. That means that even if you are at an unknown place, there has to be a probability of 1....
It's annoying about machine learning.
In fact, you might normally think that if the probabilities are all too close to each other, then that might actually indicate it is an unknown place.
Unfortunately, there is no definite way to consider what is an unknown location. It is in fact completely arbitrary from the classifier's perspective.
Only on the level of completely different buildings (when there are no known access points available), then you could build in a rule that you are most likely certain you are at an unknown location.
In a smaller area though, with overlapping access points, it is certainly not straightforward to implement "unknown location".
Is there a use case you have in mind?
The simplest case that could be implemented is checking whether the input contains only unknown wifis....
In the end, I think it only makes sense to implement this when we have use case, since there is not one obvious way to do it :)
from whereami.
@kootenpv I have problems thinking of use cases in general, so no, I do not. But I like the idea and accessibility of this project too much to just ignore it.
Maybe at a minimum it could remind you to record data (or do so itself with some daemon) if you're somewhere new?
from whereami.
Maybe this problem can be solved applying some of the techniques described in scikit-lean outlier detection
from whereami.
@fedekau Outlier detection will not help decide what the definition is of an unknown location.
Give these a thought:
- Is couch 1 a different location than couch 2?
- Is another room in the same house an unknown location or does it still count as the same location?
- Is a building the same location?"
There is no automated way to detect these changes unless you have a particular case in mind.
from whereami.
@kootenpv Yeah you're right.
The phrase unless you have a particular case
made me think, maybe a "solution" could be having an interface that the user can implement that tells the program how to define "unknow" for the specific use case they need.
By the way, awesome library 😎
from whereami.
@fedekau Thanks! Yea, you also made me rethink it like that.... But solving that right now is trying to come up with a solution without a real problem ^^
from whereami.
Awesome! I will try it 😎
from whereami.
whereami learn -l bedroom -l 100
when i execute this command the error i get is
Can't write PID file to /var/run/whereami.started.719 at /usr/sbin/whereami line 229.
can anyone explain asap
from whereami.
@vansh215 can you make sure that the whereami you use is from python? on the command line: whereis whereami
from whereami.
whereami: /usr/sbin/whereami /etc/whereami /usr/share/whereami /usr/share/man/man8/whereami.8.gz
sir,this is the output
from whereami.
@vansh215 that's not my python package, I think there's one linux distro that includes another whereami
already.
You could instead try running like:
python -m whereami learn -l bedroom # default python
python3 -m whereami learn -l bedroom # python 3, whichever you have installed
from whereami.
Related Issues (20)
- Does not refresh locations between each sample (Win 10) HOT 4
- ValueError: Valid presets for class_weight include "auto" and "subsample". Given "balanced". HOT 1
- nonfunctional due to alternative wifi device name HOT 2
- Don't show stacktrace when using Ctrl-C
- missing documentation for mac 10.11.6
- add a time dimension to locations?
- Reset knowledge HOT 1
- Update dependencies
- UnicodeDecodeError when learning HOT 1
- ValueError: n_splits=10 cannot be greater than the number of members in each class. HOT 4
- Locations lose trailing 't' or 'x' HOT 1
- How can I track my phone? HOT 8
- Does deleting {{location}}.txt files affect the model? HOT 1
- whereami rename old_label new_label not working HOT 2
- Unable to run whereami from a launch agent HOT 1
- Installing whereami on MacBook with M1 Chip running Big Sur HOT 2
- Please cooperate with the Openhab project to make binding.
- Support for different floors, aka 3D triangulation. HOT 1
- De HOT 1
- Duplicates of options still not working. When we add options and hit "save" and then the page return with 2x options
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from whereami.