619 karma · joined November 20, 2013
To alleviate this, one can add more abstract/heuristic information about the position to the input (indicators for complex relations between several pieces). This kind of high-dimensional vector would be more robust to small changes, and make the objective function more smooth. Perhaps the non-linearities introduced by the three layers cannot do this as effectively.
It's interesting to note that the brain does not have separate components for memory and computation. Every neuron computes and stores at the same time.
For me, it is most important to note what the article touches only at the end. If you have good features and representation, a linear model may be enough. Real world data and tasks usually don't require knot untying. In my opinion, the model (e.g., neural net) should be kept simple, and the challenge is in finding the right features and representation.
Coates & Ng (2012) wrote:
Many algorithms are available to learn deep hierarchies of features from unlabeled data, especially images. In many cases, these algorithms involve multi-layered networks of features (e.g., neural net- works) that are sometimes tricky to train and tune and are difficult to scale up to many machines effectively. Recently, it has been found that K-means clustering can be used as a fast alternative training method. The main advantage of this approach is that it is very fast and easily implemented at large scale. On the other hand, employing this method in practice is not completely trivial: K-means has several limitations, and care must be taken to combine the right ingredients to get the system to work well.
http://www.stanford.edu/~acoates/papers/coatesng_nntot2012.p...
A world without software patents is a world of free and faster innovation, of equal opportunity, and of open and cheap access to technology.
Firms that have patentable knowledge are strong enough to not need governmental protection. Inventing something before others gives one technical advantage and time to make profit ahead of others.
Too much effort is spent on the patent process, that could be spent directly on innovation.
In today's software world, 20 years of monopoly is like 100 years of monopoly when the constitution was signed.
The solution should probably come from the legislative branch.
The solution is to make laws that forbid software patents. And if the old patents can be abolished, the better.
> Unlike organic debris, which biodegrades, the photodegraded plastic disintegrates into ever smaller pieces while remaining a polymer. This process continues down to the molecular level.[26] As the plastic flotsam photodegrades into smaller and smaller pieces, it concentrates in the upper water column. As it disintegrates, the plastic ultimately becomes small enough to be ingested by aquatic organisms that reside near the ocean's surface. In this way, plastic may become concentrated in neuston, thereby entering the food chain.
[1] http://www.historynet.com/the-10-greatest-emergency-landings...
Texas Appeals-Court Rules Phone-Search After Arrest Violates 4th Amendment [pdf]
* The Blue Death, http://www.amazon.com/The-Blue-Death-Disease-Disaster/dp/B00...
* Principles of Neural Science (a canonical textbook)
* Immunobiology
* Genome
At issue is an update to Google's privacy policy that went into effect on March 1, 2012. The revamp consolidated 70 or so privacy policies across Google's products down to one. But with this change, Google also switched to one profile for users across all services rather than separate logins for offerings like YouTube, Search, and Blogger.
* A site may be able to compromise the browser extension.
* You have to memorize several passwords because sites require different length passwords.
* The code has been reviewed less.
* A key-store like KeePass can store many original passwords, not just one hashed password.
* It doesn't have a non-browser app, so I had to copy paste passwords from the browser, while KeePass has Alt-Ctrl-A.