[R] Google, Cambridge, DeepMind & Alan Turing Institute’s ‘Performer’ Transformer Slashes Compute Costs Computer Science |
- [R] Google, Cambridge, DeepMind & Alan Turing Institute’s ‘Performer’ Transformer Slashes Compute Costs
- What are some educational content made by companies?
- Number of maximal cliques in a (2C_4, C_5, P_5)-free graph
- [R] Google AI ‘Looking to Listen’ Tech Enables Speech Enhancement on YouTube Stories in Seconds
- MY issues with The Halting Problem
- A Novel Approach to Generate Correctly Rounded Math Libraries for New Floating Point Representations
- Regarding sources of any kind for going in depth in Genetic Algorithms
- MS Access Alternative for Ubuntu
- Microsoft recently released an open source AI alongside a paper explaining how it works.
| Posted: 02 Oct 2020 01:18 PM PDT A team from Google, University of Cambridge, DeepMind, and Alan Turing Institute have proposed a new type of Transformer dubbed Performer, based on a Fast Attention Via positive Orthogonal Random features (FAVOR+) backbone mechanism. The team designed Performer to be "capable of provably accurate and practical estimation of regular (softmax) full rank attention, but of only linear space and timely complexity and not relying on any priors such as sparsity or low-rankness." Here is a quick read: Google, Cambridge, DeepMind & Alan Turing Institute's 'Performer' Transformer Slashes Compute Costs The paper Rethinking Attention With Performers is on arXiv. [link] [comments]  |
| What are some educational content made by companies? Posted: 02 Oct 2020 10:07 PM PDT I recently discovered The Amazon Builder's Library where they write informative articles about software challenges Amazon faced when scaling up. Previously, I only knew about Riot's So You Wanna Make Games game development series on YouTube. Generally, educational articles/content provided by companies are more practical and go more in-depth compared to other articles online. What are some other educational content made by companies? [link] [comments]  |
| Number of maximal cliques in a (2C_4, C_5, P_5)-free graph Posted: 02 Oct 2020 08:12 PM PDT So far, I have found out that chordal graphs have linear number of maximal cliques with respect to the number of vertices. In general case, it is exponential. I am trying to determine whether the number of maximal cliques in a (2C_4, C_5,P_5)-free graph with respect to the number of vertices. In a (2C_4, C_5,P_5)-free graph, the largest induced cycle is of length 4, and no two 4-cycles are edge-disjoint i.e. there always exists an edge which connects two 4-cycles. Is there a paper that mentions such result? [link] [comments]  |
| [R] Google AI ‘Looking to Listen’ Tech Enables Speech Enhancement on YouTube Stories in Seconds Posted: 02 Oct 2020 08:18 PM PDT Google AI has announced a new audiovisual speech enhancement feature in YouTube Stories (iOS) that enables creators to make better selfie videos by automatically enhancing their voices and reducing noise. The new feature is based on Google's Looking to Listen machine learning (ML) technology, which uses both visual and audio cues for isolating and separating the speech of a video subject from background sounds. Here is a quick read: Google AI 'Looking to Listen' Tech Enables Speech Enhancement on YouTube Stories in Seconds The paper Looking to Listen at the Cocktail Party: A Speaker-Independent Audio-Visual Model for Speech Separation is on arXiv. [link] [comments]  |
| MY issues with The Halting Problem Posted: 03 Oct 2020 02:25 AM PDT So, I've had this personal bee in my bonnet for close to a decade. The proof by contradiction solution to the halting problem, in my opinion makes a few blatantly obvious errors and it's shocked me that anyone takes that proof seriously at all. The Halting Problem is the question whether it would ever be possible to tell if an algorithm or program would halt, or continue running forever, using another program/algorithm. Now with some very basic logic we can decide a couple obvious ground rules: I make a program the loops forever until an input is given and then it halts. Does this program halt, or run forever? Obviously that question cannot be reasonably asked, as it obviously hinges on the input.So either the input must be defined and included as part of the program, all inputs pre decided, OR we include a possibility for there to be a third answer where the program will only halt if given X range of inputs. If I make a program that is infinitely long, there will never be a reasonable way to determine its outcome, as there will always be more code that it could jump to. So with these common sense rules in place, let's take a look at this age old "proof"... Given a program H(P, I) that returns True if the program P, halts given input I, and returns False if p will never halt. so far so good, I statically defines the input to P, in this description of the proof if we define a program Z as: Now we're starting to see something off, in this case P takes itself as an input, not entirely bad, you could chuck a whole program in the standard in, Consider what happens when the program Z is run with input Z See, now this is where this jumps the rails, because let's take a look at this Z(Z) Z(Z) = if (H(Z,Z)) { while(true); } else { break; } As we can see, by passing in Z recursively you end up with a functionally infinite program, purposely intended to have each rung down alternate between halting or looping, causing a paradox. And as we pointed out before an infinite program is fundamentally unsolvable regardless of any absurd paradoxes. This whole proof ends up feeling like someone saying "you can't divide by 0, therefore division can't exist". Obviously division works perfectly fine when you understand that behaviours involving or resulting in infinity, like dividing by 0, are often invalid or undefinable. Likewise the halting problem should still be solvable provided you understand that any code must be finite in length. Thankyou for coming to my rant (this ain't a TED talk, this is me yelling at a dead guy for an hour straight) [link] [comments]  |
| A Novel Approach to Generate Correctly Rounded Math Libraries for New Floating Point Representations Posted: 02 Oct 2020 08:31 AM PDT |
| Regarding sources of any kind for going in depth in Genetic Algorithms Posted: 02 Oct 2020 10:37 AM PDT I would have just googled this but what I know about this field is it is a highly active research area, so much of the stuff gets outdated and it is good to for newcomers to start with the latest edition books or newly created blogs for elementary people etc. So that is why I intend to ask the people of this community regarding the approach to take a deep dive in the field of Genetic Algorithms. Recommendation of any form such as online course, standard textbook, informative blogs etc will do! Also if someone also gives the method to approach this field, then I will be highly obliged! [link] [comments]  |
| MS Access Alternative for Ubuntu Posted: 02 Oct 2020 09:45 AM PDT Hi. We are studying intro to databases, and we are using MS Access. I want to know some of the alternatives which provide the same functionality as MS Access. Libre Base was missing some functionalities such as Validation rules and text. Suggest some [link] [comments]  |
| Microsoft recently released an open source AI alongside a paper explaining how it works. Posted: 02 Oct 2020 12:43 PM PDT |
| You are subscribed to email updates from Computer Science: Theory and Application. To stop receiving these emails, you may unsubscribe now. | Email delivery powered by Google |
| Google, 1600 Amphitheatre Parkway, Mountain View, CA 94043, United States | |
No comments:
Post a Comment