How can beginners in machine learning, who have finished their MOOCs in machine learning and deep learning, take it to the next level and get to the point of being able to read research papers & productively contribute in an industry?
Try to build machine learning systems and achieve close to state of the art performance on benchmark data set. For example, build a Convolutional Neural Network and apply it on Street View House Number data set. As for reading papers, it may be a good idea to replicate a few papers. Hope this helps.
List of important projects and papers in part 3 of the tutorial given at
https://adeshpande3.github.io/adeshpande3.github.io/A-Beginner%27s-Guide-To-Understanding-Convolutional-Neural-Networks/
I always go to my community (city, municipality), sit on professional meetings, get out of my comfort zone and visit other fields. All of a sudden you will be presented with a ton of problems that can benefit from your experience as a ML practitioner -- obviously your role would be to offer solutions to their problems -- this way i ran across designing an anomaly detection system in irrigation channels, designing computer vision system for trash composition analysis at the local trash transfer station, predicting power generation from renewable energy at islanded microdgrids...... and the list keeps going on an on.
From my point of view, the first thing one has to do is to decide what excites him/her. The machine learning area is so broad, covering many different topics, and even if you would like to learn all of these, you can not do it in an efficient way ( just scratch the surface of each problem). If you are not yet there, I would go with Mengfei’s recommendation.
Let’s say that the research field you like to read and contribute to is hypothetical reasoning on language generation. Then by reading the papers’ abstract you should be able to know if you could grasp the basic idea. In the beginning of reading others research, you should expect that you will not understand every part of it; but you would certainly understand what is their contribution area and with what approaches/techniques they achieve that.
As far as industry concerns, again it depends on what your role want to be (e.g. an NLP algorithms engineer or data scientist on applying image recognition techniques). After that, I would go with participating in Kaggle competitions and trying to understand the problem and data in hand (instead of just doing benchmarking of models towards prediction accuracy:’If it predicts it works’ is not a solid statement). At the same time I would choose a book with practical examples (or a well written thesis) and try to follow it.
Hope that helps you
Please look at these Github repository:
https://github.com/kk7nc/RMDL
https://github.com/kk7nc/HDLTex
https://github.com/BarnesLab/Patient2Vec
Please read following paper:
Conference Paper RMDL: Random Multimodel Deep Learning for Classification
Conference Paper HDLTex: Hierarchical Deep Learning for Text Classification
Article An Improvement of Data Classification using Random Multimode...
Conference Paper Analysis of Railway Accidents' Narratives Using Deep Learning
'Article Patient2Vec: A Personalized Interpretable Deep Representatio...
- Badges
- Science topic