Google Tech Talk
Presented by Yoshua Bengio
Yoshua Bengio will give an introduction to the area of Deep Learning, to which he has been one of the leading contributors. It is aimed at learning representations of data, at multiple levels of abstraction. Current machine learning algorithms are highly dependent on feature engineering (manual design of the representation fed as input to a learner), and it would be of high practical value to design algorithms that can do good feature learning. The ideal features are disentangling the unknown underlying factors that generated the data. It has been shown both through theoretical arguments and empirical studies that deep architectures can generalize better than too shallow ones. Since a 2006 breakthrough, a variety of learning algorithms have been proposed for deep learning and feature learning, mostly based on unsupervised learning of representations, often by stacking single-level learning algorithms. Several of these algorithms are based on probabilistic models but interesting challenges arise to handle the intractability of the likelihood itself, and alternatives to maximum likelihoods have been successfully explored, including criteria based on purely geometric intutions about manifolds and the concentration of probability mass that characterize many real-world learning tasks. Representation-learning algorithms are being applied to many tasks in computer vision, natural language processing, speech recognition and computational advertisement, and have won several international machine learning competitions, in particular thanks to their ability for transfer learning, i.e., to generalize to new settings and classes.
PhD in CS from from McGill University, Canada, 1991, in the areas of HMMs, recurrent and convolutional neural networks, and speech recognition. Post-doc 1991-1992 at MIT with Michael Jordan. Post-doc 1992-1993 at Bell Labs with Larry Jackel, Yann LeCun, Vladimir Vapnik. Professor at U. Montreal (CS & operations research) since 1993. Canada Research Chair in Statistical Learning Algorithms since 2000. Fellow of the Canadian Institute of Advanced Research since 2005. NSERC industrial chair since 2006. Co-organizer of the Learning Workshop since 1998. NIPS Program Chair in 2008, NIPS General Chair in 2009. Urgel-Archambault Prize in 2009. Fellow of CIRANO. Current or previous associate/action editor for Journal of Machine Learning Research, IEEE Transactions on Neural Networks, Foundations and Trends in Machine Learning, Computational Intelligence, Machine Learning. Author of 2 books and over 200 scientific papers, with over 9000 Google Scholar citations in 2011.
This site provides links to random videos hosted at YouTube, with the emphasis on random.
The original idea for this site actually stemmed from another idea to provide a way of benchmarking the popularity of a video against the general population of YouTube videos. There are probably sites that do this by now, but there wasn’t when we started out. Anyway, in order to figure out how popular any one video is, you need a pretty large sample of videos to rank it against. The challenge is that the sample needs to be very random in order to properly rank a video and YouTube doesn’t appear to provide a way to obtain large numbers of random video IDs.
Alternative random YouTube videos generator: YouTuBeRandom
Even if you search on YouTube for a random string, the set of results that will be returned will still be based on popularity, so if you’re using this approach to build up your sample, you’re already in trouble. It turns out there is a multitude of ways in which the YouTube search function makes it very difficult to retrieve truly random results.
So how can we provide truly random links to YouTube videos? It turns out that the YouTube programming interface (API) provides additional functions that allow the discovery of videos which, with the right approach, are much more random. Using a number of tricks, combined some subtle manipulation of the space-time fabric, we have managed to create a process that yields something very close to 100% random links to YouTube videos.
YouTube is an American video-sharing website headquartered in San Bruno, California. YouTube allows users to upload, view, rate, share, add to playlists, report, comment on videos, and subscribe to other users. It offers a wide variety of user-generated and corporate media videos. Available content includes video clips, TV show clips, music videos, short and documentary films, audio recordings, movie trailers, live streams, and other content such as video blogging, short original videos, and educational videos. Most content on YouTube is uploaded by individuals, but media corporations including CBS, the BBC, Vevo, and Hulu offer some of their material via YouTube as part of the YouTube partnership program. Unregistered users can only watch videos on the site, while registered users are permitted to upload an unlimited number of videos and add comments to videos. Videos deemed potentially inappropriate are available only to registered users affirming themselves to be at least 18 years old.
YouTube and selected creators earn advertising revenue from Google AdSense, a program which targets ads according to site content and audience. The vast majority of its videos are free to view, but there are exceptions, including subscription-based premium channels, film rentals, as well as YouTube Music and YouTube Premium, subscription services respectively offering premium and ad-free music streaming, and ad-free access to all content, including exclusive content commissioned from notable personalities. As of February 2017, there were more than 400 hours of content uploaded to YouTube each minute, and one billion hours of content being watched on YouTube every day. As of August 2018, the website is ranked as the second-most popular site in the world, according to Alexa Internet, just behind Google. As of May 2019, more than 500 hours of video content are uploaded to YouTube every minute.
YouTube has faced criticism over aspects of its operations, including its handling of copyrighted content contained within uploaded videos, its recommendation algorithms perpetuating videos that promote conspiracy theories and falsehoods, hosting videos ostensibly targeting children but containing violent and/or sexually suggestive content involving popular characters, videos of minors attracting pedophilic activities in their comment sections, and fluctuating policies on the types of content that is eligible to be monetized with advertising.