Q1: How often do you read a research paper discussing a deep learning algorithm ?
Q2: Assuming that you come across a relevant research paper, how much time does it take for you to understand the research paper?
Q3: Assuming that you are successful in understanding the paper, how much time does it take for you to implement the paper in a single library (say Tensorflow) ?
Q4: Assuming that you implement the research paper in a single library, how much time does it take for you to implement the research paper in other libraries (say Keras, Caffe, PyTorch etc) ?
Q5: Assuming that you implemented the research paper in multiple libraries, how much time does it take for you to debug the implementation and make it bug free?
Answer: Assuming that you have read this post till here, what if I say that the time taken for you to read the above 5 questions is now more than the time taken for you to perform them!
This is our goal in our research paper published in the AAAI Conference on Artificial Intelligence (AAAI-18) called “DLPaper2Code: Auto-generation of Code from Deep Learning Research Papers”
Please find the poster here, that we presented in AAAI-2018 conference.
Please note that this feature is an initial proof of concept, and is not yet available as a system for open consumption.
IBM Deep Learning IDE (DL-IDE)
Along with this, we have built our easy-to-use, intuitive IDE for deep learning called -IBM Deep Learning IDE (DL-IDE). Play around with it - lots of cool features secretly built into it and lot more coming!
Watch the demo video here for the existing features and features to come -
AAAI 2018 Presence
I am presenting four papers (1 AAAI, 1 IAAI, 2 Technical Demo) in #AAAI2018, New Orleans (AAAI)
For more information on the papers, visit IBM Research AI Blog
If you are around, do reach out to me at email@example.com or tweet to me @goodboyanush