Finnish Center for Artificial Intelligence FCAI (Suomen tekoälykeskus) is a community of experts that brings together top talents in academia, industry and public sector to solve real-life problems using both existing and novel AI.
FCAI is one of the Research Council of Finland Finnish flagships, hubs of top-level research and impact.
Really interesting and great see a vibrant action.. 1st hand human testimony credibility definitely goes back up it calls for a return of the international journalist membership standardized 3 sources confirmed or it's just not news for those journalist. Archives of older conventional ways of doing things can deal with much of the coming challenges. Similar with horsepower robot slaves and utility serfs of national gdp. I see no reason to become welfare states when a past aristocracy could author history be private scientist inder Similar Free power and utility lol Maybe even stock bestowed at birth legal citizen ownership helps avoid weird movements to free bots with voting rights lol Europe is far removed from these older methods of life. Usa Is just 120 yrs of reversal since esoterica America founders had foresight on the future beast of burden to come. Mature society up some assume individual have a gun holster on the hip loaded with whatever nature permits mutual destruction created shilvery and respect with form in time. Last time individuals had such liberty pursuit freedom improving the human condition was launched. Strong moral law & courts and surly family cells will be incentivised more and more than anything 1900s structuralism could provide.
Yes, LIME and SHAP methods are model-agnostic. You can use them on what ever deeplearning model you want, but the explanation which these methods are giving is unfortunatelly model and data spesific.
Fantastic talk, and a really good paper. Over 400 citations in two years... a lot to sift through! Would love to see a follow up! I need to dive into the guidance next, this just leaves me wondering why we don't use clip to guide the diffusion of a low resolution semantic map first, that captures the structure, meaning, long range patterns easily, maybe even a course depth map too, and then use that to guide the diffusion of a latent, which then gets intelligently upscaled guided by the semantics.
Thanks for the great talk! I wonder whether there is any inherent reason why you couldn't do the optimization that you do on the prior on a variational posterior directly or use it somehow in HMC directly. The chaining of two approximations seems tricky, but I guess an approximately functional prior is still better than the standard non-functional priors.
I d like to point out something i disagree with : the action space isnt actually 4x18x2, its 4*(18x32)(number of units, in this case cards in your hand, to the power of the possible actions number). Let me explain. A game mechanic not mentioned in the video is that the player can actually deploy multiple cards in the arena at once, even up to all 4 cards from you hand. Thats a game changer. I know that this mechanic is most of the time useless, but in order to buid a theoretically "perfect" agent we have to take into account every single possibility. Most of the game is still umexplored(wich is logical considering the fact that Clash Royale is only 6 years old) and new ways of playing the game are discovered all the time. Recently i have seen more and more tactics involving simulteanously placing two cards in the arena. The agent should explore the possibilities that arise with the use og such tactics. Only then it will be complete. I also want to mention that i dont have specific knoledge on the subject, so i could be wrong. Its just my opinion. Have a nice day!
In 17:31, You talked about p(noisy_pixel / clean_pixel, noisy_contex), but since noise is pixel-wise independent, isn't this just equal to p(noisy_pixel/clean_pixel). Because intuitively, learning about the context introduces no new information since we already knew the clean value.
This guy is in charge of something? Sometimes getting an education and being practical in the field are way way off from each other, I always respected in Finnish Center for Artificial Intelligence FCAI until this talk, I’m gonna pretend it never happened
11:21 is that true?? someone needs to step-up and help this guy talk, not every mathematician can talk to an audience and alot of times even can't make sense when talking at all because the mind is going to fast
Very fantastic work. I am seeking some efficient way to approximate Bayesian posterior distribution except MCMC. Your research inspired me a lot! I am just curious what do you mean 'not-too-complex' posterior distribution. Thanks!
