Timeline for Why do neural networks need so many training examples to perform?
Current License: CC BY-SA 4.0
3 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 28, 2019 at 21:16 | history | edited | Scholar | CC BY-SA 4.0 |
edited body
|
| Feb 28, 2019 at 20:37 | comment | added | corey979 | I would add also that the child has context: it sees a train on the rails, be it at a station, level crossing etc. If it sees a huge (zeppelin size) balloon shaped and painted to look like a train in the sky, it won't say it's a train. It will say it looks like a train, but it won't attach a label "train" to it. I'm skeptical a NN will return a label "train-looking balloon" in this case. Similarly, a child won't mistake a billboard with a train on it with an actual train. A picture of a picture of a train is a picture of a train to a NN – it will return the label "train". | |
| Feb 28, 2019 at 19:15 | history | answered | Scholar | CC BY-SA 4.0 |