Talk:Conflict-driven clause learning

---

---

Overall Quality of the Article

I just stumbled over this article and, observing some obvious flaws, incorporated some slight changes. After this, I had to note that the quality of the whole article is "suboptimal". I think that the whole article should be revised. Ben Chumin 16:26, 21 April 2016 (UTC)

Flaws remain.

In is not made explicit that CDCL is DPLL augmented.

The Example uses a different notation than that introduced at the beginning (+ instead of \/, etc).

"3.Build the implication graph." -- it is not explained what implication graph is.

(Well, it is kind of explained in the example, and a link is given). I am adding the link at the first appearance.

Starting from Step 6 in the example, the label of one node is unreadable.

The diagrams Step 7 and Step 8 seem the same.

Wlodr (talk) 17:05, 26 November 2022 (UTC)

Press release?

This page just looks like someone's spam press release, especially with its unreferenced claims for its superiority to the standard algorithms.

CDCL is currently (as of 2014) the standard family of algorithms for solving SAT. David.Monniaux (talk) 00:18, 5 December 2014 (UTC)

Probably, you are right; but we'd need a source for this. Isn't there an introductory section in an overview article saying this? - Jochen Burghardt (talk) 07:49, 5 December 2014 (UTC)

Tree resolution vs DAG/linear resolution

Shouldn't the difference between DPLL (no learning) and CDCL also be explained in terms of tree resolution (no sharing of subproofs) vs DAG/linear resolution (learning of lemmas, thus shared subproofs)? David.Monniaux (talk) 20:19, 27 July 2014 (UTC)

I would like to have such a section. However, before that, the CDCL approach should be explained at all - the current explanation is very poor, in particular item 4. Thanks god there is a good example following. - Jochen Burghardt (talk) 07:58, 5 December 2014 (UTC)

Efficient Algorithm

Claiming that this algorithm is "efficient" es misleading since the usual definition of efficiency is that the algorithm runs in polynomial time, which is not the case (especially given that SAT is NP-complete) — Preceding unsigned comment added by 146.155.23.42 (talk) 21:00, 22 September 2014 (UTC)

Lead as overview

There needs to be at least one sentence in the lead that depicts the philosophy or approach of the algorithm. WalkSAT is stochastic (low retention of decision history). Surely this is something. Backtracking? — MaxEnt 18:01, 16 April 2018 (UTC)

Unique implication point

a lot of the literature talks about UIPs but the example skips this. Is the conflict a UIP as well? 94.173.239.33 (talk) 20:16, 16 May 2023 (UTC)

Cut not defined?

Hi, the article talks about finding the "cut", but this term is not formally defined. It would be good to add a formal definition. --a3nm (talk) 23:11, 7 November 2023 (UTC)