2.75

Is it known what they use for the synthetic training data? From my read I'd guess they construct 'random' causal graphs w/ hyperparameters so that the distribution of synthetic datasets implied by those graphs somehow resembles the distn of relationships between variables in some reference datasets

Now this is the real "students these days"

Some claim it's Tokio Hotel

👋

...one particular mark, more like

Great talk!

Is this residualized (in a weird way)? Don't see how NE would be lower than Midwest in absolute terms otherwise

Neat! Any plans (or thoughts on feasibility) to add some RAG/semantic search capabilities to it? Like, "Papers that use two-way fixed effects"?

"Danke" 🧑‍🍳😘👌

Model me this, DAG people

I recently got into using darktable for (RAW) photos and have been pretty happy with it so far

Yeah I read OP as making a statement about existence of the expected value

On a somewhat related note, I always wondered why it seemed* to me that the French hated the collaborators with greater intensity than the Nazis themselves. I now have an idea why

*(Not a historian)

Strangely, it is hence the only brow that has lost all respectability

They accurately quote a wrong source, so it remains wrong :)

The *value* of Canada's economy = value of all Canadian firms combined.

The *production* (or output) of Canada's economy = Canadian GDP.

Author says the first and refers to second.

Also doesn't make logical sense btw - compare NVDA value to CAN value (or NVDA prodn with CAN prodn) instead