@emollick The harder problem was category granularity.
Fine-grained task categories improved MAB performance because tasks were more comparable within each bandit. But the trade off was data fragmentation: more bandits, fewer samples per bandit.
@emollick Agree. I tested a multi arm bandit project using e-greedy with weights tied to user specified kpis.
Findings after testing were that the regret cost for exploration was too high.
I haven't tested a more sophisticated selection policy though.
@GBNEWS Fucking shite idea.
Read Orwell's down and out in Paris and London for an intro on how the meaning and weight of swear words change over time. Language is class coded, the offensiveness of a word isn't fixed.
@RokoMijic Look, I tend to agree. Regardless of the sci-fi physics, dark forest theory should be be considered in game theoretic expansion.
The third book alludes to a hiding nature, prey like behaviour.
If we go for expansion (in our nature), how do we morally encounter the prey?
@tomhfh I used to pay a ridiculous sum to rent a flat in Canary Wharf. Specifically, the new development, Wood Wharf.
Pros, my fiancé could walk to the shop and feel safe.
Cons, largely soulless.
@PeterSchiff It's the fixed supply that matters, not how the pieces are denominated.
You can split a dollar into cents, but a dollar is a dollar.
Increasing the divisibility of Btc even beyond Satoshis serves liquidity and does not generally harm scarcity.