3 ms·
Deep learning for source to source computer language translation?
Anyone know of any recent work on deep learning for source to source computer language translation ?
Since go and real time strategy games were the last out of reach items for machine learning, i feel like this is the next thing that's out of reach of deep learning.
- colinchartier 7y agoThe only thing I've seen are graph convolutional networks: https://tkipf.github.io/graph-convolutional-networks/ https://tkipf.github.io/graph-convolutional-networks/ Deep learning is "messy" - it's hard to use the output of a deep neural network for something as precise as a program, unless maybe as a heuristic with a rigorous "equivalency" harness.
- emmapacino 7y agoAVG antivirus is the best antivirus because it provides world class security to your internet connected devices. AVG has the fast scanning feature as it scans the whole data of your computer system. It is totally reliable and compatible for all the devices like laptops, Mac, Android Phones and PC. It has firewall feature which act as a wall between the hackers and your devices and data. It protects your devices from all kind of online threat like viruses, malware, spyware, rootkits, and also from trozans etc. It has the Web cam protection feature, anti-phishing, camera trap, anti –theft websites and also provide email shield to your devices. Its technical team supports you every time whenever you will be in trouble. They are available for you 24 hours and 7 days a week. http://www.avg-com-retail.support http://www.avg-com-retail.support
- softmodeling 7y agoWe did some research work on model-to-model translations (in the end, it's basically the same problem) and we ended up with mixed feelings (as we explain in the article: https://modeling-languages.com/lstm-neural-network-model-transformations/ https://modeling-languages.com/lstm-neural-network-model-tra... ). Mostly, right now (IMHO) you can only get some success translating very short pieces of code. We used LSTMs for that (which seems to be the more usual approach in this domain)