Deep learning algorithms are prone to a previously unknown problem, say a team of computer scientists at Google.
https://www.discovermagazine.com/technology/google-reveals-major-hidden-weakness-in-machine-learning
I think such difficulties are mainly due to very little understanding mathematical backgrounds for the problems at hand. Basic need is to understand mathematical/statistical model based learning within specified domains.
I think such difficulties are mainly due to very little understanding mathematical backgrounds for the problems at hand. Basic need is to understand mathematical/statistical model based learning within specified domains.
Earlier this year Dr. Rhodes' wrote the linked article, which discusses assessing outcome reproducibility using multiple (random) seed values:
- https://odajournal.com/2020/06/23/assessing-reproducibility-of-novometric-bootstrap-confidence-interval-analysis-using-multiple-seed-numbers-invited/
Last year Dr. Linden reported that false positives ARE a problem for various statistical models adopted in machine-learning algorithms:
- https://odajournal.com/2019/03/21/some-machine-learning-algorithms-find-relationships-between-variables-when-none-exist-cta-doesnt/
- https://odajournal.com/2019/05/16/logistic-discriminant-analysis-and-structural-equation-modeling-both-identify-effects-in-random-data/
- https://odajournal.com/2019/04/22/multi-layer-perceptron-neural-net-model-identifies-effect-in-random-data/
- https://odajournal.com/2019/04/11/effect-of-sample-size-on-discovery-of-relationships-in-random-data-by-classification-algorithms/
This has been known to the ML community for a long time, just that nobody thought to state it as a problem
- Badges
- Science topic