2 ms·
The paper is actually pretty interesting - it's a pity the headline is misleading. The Reddit comment[1] is a good summary: More accurately: "The Tsetlin Machi
by nl 8y ago
The paper is actually pretty interesting - it's a pity the headline is misleading.
The Reddit comment[1] is a good summary: More accurately: "The Tsetlin Machine - a new approach to ML - outperforms single layer Neural Networks, SVMs, Random Forests, the Naive Bayes Classifier and Logistic Regression on four carefully selected contrived datasets"
Notably, on their (weird!) "Binary Iris" dataset, the NN appears to be undertrained, and it is unclear what the headline "mean" accuracy figure is actually a mean of.
However, once we get over that, it's quite a different approach, and I can imagine places where it could be useful. Notably, as an anomaly detector it would seem to have interesting properties like interpretable results similar to a random forest.
[1] https://www.reddit.com/r/MachineLearning/comments/89yp8g/r_the_tsetlin_machine_a_new_approach_to_ml/dwujoxt/ https://www.reddit.com/r/MachineLearning/comments/89yp8g/r_t...
- nestorD 8y agoThey say that their results are interpretable due to being logical formulas but, later, show a toy problem in which their formula has 10000 clauses. By that definition deep-learning is probably interpretable. However I do think that it is an interesting concept that would probably be worth testing if your inputs and outputs have a natural mapping to booleans.