6 Apr 2024 |
| lusko set a profile picture. | 17:55:01 |
8 Apr 2024 |
| @azraelsdad:matrix.org joined the room. | 07:45:39 |
| @azraelsdad:matrix.org left the room. | 07:46:55 |
| antogiro joined the room. | 19:51:58 |
9 Apr 2024 |
tactt | In reply to @tcsvn:matrix.org Short Question: what is the proper/currently best performing way to treat an "other" class in supervised classification. Assuming you classify images with 30 classes. The "other" class would be everything but those 30 classes. One way would be to use uncertainty of the softmax. Another would be to create an additional 31th class that ist filled with "everything" else.
There is suprising little information on how ML systems in production tackle this issue. Thanks already for your replies☺️ It seems that the first way means to forcibly divide the "other" class into the 30 classes with uncertain probabilities, while the second way creates a new class. In my opinion, there are two approaches to solve it, as you said, maintain the uncertainty or create another class. You can set a default value or use a nearest neighbor algorithm (if possible) for strange cases or co-variables, or directly create a new class to indicate that those that cannot be classified with the known classes would also be accepted. You can try both and then choose the more practical one. | 03:44:04 |
| AtomLaw joined the room. | 06:23:47 |
| ali111111 joined the room. | 09:46:14 |
| ayakasakura joined the room. | 21:52:59 |
ayakasakura | Oh man, I got hacked! But hey, I made it into a song.
https://sakuraayaka.bandcamp.com/track/-
I've been turning the real-time chaos captured on my phone and computer into songs to spread the word. I'm keeping this up, turning each real moment into music. This is the face of national censorship in Japan, but I'm not backing down. Let's make this go viral! | 21:53:05 |
11 Apr 2024 |
tactt | Please do not send unrelated messages. This room is for data science and machine learning. ayakasakura | 13:14:52 |
12 Apr 2024 |
| @fromtheeast:matrix.org left the room. | 13:10:07 |
13 Apr 2024 |
| @abhiraj_karki:matrix.org joined the room. | 09:10:39 |
| @abhiraj_karki:matrix.org left the room. | 10:38:15 |
| @puppetic:matrix.org joined the room. | 14:43:05 |
15 Apr 2024 |
| netguru joined the room. | 12:02:55 |
16 Apr 2024 |
| @puppetic:matrix.org left the room. | 06:02:06 |
| ali111111 set a profile picture. | 07:01:51 |
| unicorny9205 joined the room. | 08:53:49 |
| ali111111 changed their profile picture. | 09:49:37 |
| abraarsyed joined the room. | 09:52:42 |
| ali111111 changed their profile picture. | 09:58:17 |
| jhuc joined the room. | 19:41:36 |
17 Apr 2024 |
| astrotee joined the room. | 10:43:29 |
thompson_plug | Hello everyone I'm a Grower and supplier of cannabis and psychedelics products such as;
edibles
Distillates
Disposables
Shrooms(mushrooms)
Weed (bud)
Vapes
Codeine
CBD oil
Packwood
Cocaine
Ketamine
Steroids
Pills
Hash
Rosin
Mdma
DMT
Meth
Cart
Clone card
Cashapp/PayPal flip
Bank notes too and many more...
Let me know if you’re interested
Check out more products in channel 👇 👇👇
https://t.me/hightime_markert | 23:06:47 |
18 Apr 2024 |
| @thedisruptivecollective:matrix.org changed their profile picture. | 09:45:49 |
| scaryfurbies joined the room. | 13:33:20 |
| @thedisruptivecollective:matrix.org left the room. | 15:11:10 |
| @hoshizuki:matrix.org left the room. | 17:12:08 |
19 Apr 2024 |
| Torsten Osigus changed their display name from tosigus to Torsten Osigus. | 09:08:33 |
| @khaled358is:matrix.org left the room. | 19:15:55 |