Deep Learning -DL- is still one of the hottest topics today. Every company that works with lots of data is now using DL to distill knowledge from it, to become better at what they are doing and that includes the domains of audio, video, images or text. And very often, the results speak for itself. As we pointed out before, DL is hardly a new concept but with all the available data and the processing power, we can finally use the big guns to analyze the tons of data.
And that there is a demand for such services can be also seen, if we consider how many companies today have ‘Deep’ in their names, even if most of them are now part of bigger companies, which is also proof that knowledge in DL is very valuable these days. However, since DL is just a term it probably means different things to different people. Some might think of it as a synonym for big, deep, neural nets, while other might think of it as models that provide a hierarchical representation of knowledge. Of course, millions of other interpretations are also possible.
Nevertheless, regardless of what we think of the DL hype, the research that was done in the few last years provided a lot of new insights and helped to get a better understanding of the material. And with the help of sites like arxiv, the public access to high quality papers is easier than ever. Plus, not only universities, but also companies publish parts of their work there which can bee seen as further proof that DL also works very good on practical problems.
However, with the recent emerge of all the new toolkits that allow to use DL with just a few clicks, it makes it sometimes very hard, to find out if somebody is a swashbuckler or a swordsman.