Powered by NASSCOM's CoE for IoT & AI and NVIDIA
This is an internship program that will teach Machine Learning and related topics being organised by MLBLR. The program is divided into different phases and only after clearing a phase, we can move to next one. This is a fast paced program and won't be dealing with topics that are irrelevant in the industry today. Looking forward to gaining a lot from this program. Details of my progress would be updated here. The details of the program are mentioned below:
The External Internship Program is a Machine Learning for Deep Neural Networks internship program designed especially for working Professionals, though it is open for learners from all domains.
The strategy for EIP is to focus more on the HandsOn experience first, and then, take you deeper into concepts. Artificial Intelligence and Machine Learning are evolving extremely fast which makes the concepts invented last year, obsolete this year. Therefore we will cover mostly the latest concepts and not spend time on historical stuff not being used in the industry anymore.
EIP has two objectives:
-
The first objective of EIP is to unravel deeper concepts currently in use in ML algorithms. If you are able to read through a new research paper in a breeze, read it's code and reimplement it with improvements, we will consider that we have achieved our objective.
-
The second objective of EIP is to democratize AI and ML knowledge. We want to develop a strong ecosystem in India for AI/ML and this directly depends on how many people actually practice it.
EIP Version II is divided into 3 phases. You need to clear the exam at the end of each phase to progress forward. By design, only those who land in the top 70 Percentile will proceed forward.
- Phase I has 4 sessions (covering topics from Basics to CNN)
- Phase II has 4 sessions (covering Advanced CNN and Project)
- Phase III has 6 sessions (covering LSTM and Reinforcement Learning)
| Session | Topic | Prerequisite | Completion Status |
|---|---|---|---|
| 1 | ML Basics & 1st DNN HandsOn | NIL | ✔️ |
| 2 | Math & Python Foundation | NIL | ✔️ |
| 3 | Convolution Foundations | NIL | ✔️ |
| 4 | Deeper Insights | NIL | ✔️ |
| 5 | Advanced Concepts Part I & Project 1A | Phase 1 Exam and Assignment | |
| 6 | Advanced Concepts Part II & Project 1B | NIL | |
| 7 | Transfer Learning & Project 2 | NIL | |
| 8 | Embeddings | NIL | |
- |
Certification | 50 Percentile in Phase 2 Exam | |
| 9 | LSTM | 70 Percentile in the Phase 2 Exam and Project 2 | |
| 10 | GAN Part I | NIL | |
| 11 | GAN Part II | NIL | |
| 12 | Reinforcement Learning Basics | NIL | |
| 13 | RL Common Approached | NIL | |
| 14 | Policy Gradient Part I & Project 3 | NIL | |
| 15 | Policy Gradient Part II | NIL | |
| 16 | Evolution Strategies | NIL | |
- |
Advanced Certification | Successful Completion of Project 3 |
Test-driven-Approach - Every assignment is collected and marked. Every intern must have a cumulative score of more than 70% for the past 4 assignments to be eligible, without exception, for the Phase II and onwards. To start working on an individual project you'll need to have a firm grasp of the concepts covered till 4th session. Next phase focuses on much deeper concepts like complex model training, transfer learning, etc.
- Only if you are eligible for the next phase, the content will be available to you.
- Interns will be shuffled after the 4th session, and new batches will be assigned.
From the 5th session onwards, you'd be assigned individual projects which would involve training your own neural network to solve a unique problem. We plan to finish 2 projects before your internship (Phase I and II) ends.
You will be eligible for a completion certificate if you are able to finish 8 Sessions with project successfully completed, and your score is 50 percentile and above. Only if your cumulative score is more than 70 percentile, you will be eligible to proceed for the third Phase.
EIP Version II will also cover Reinforcement Learning. We have clubbed LSTM and GANs with RL as there is much to be gained if we learn them together.
To be frank, CNNs today are already commoditized and if you are looking for an edge, you must know how to write RL algorithms. Moreover, the process in CNN has sort of plateaued, and almost all older problems are being solved with RL.
Phase III will have an RL project which will challenge your understanding and help you gain a working perspective.
The Fourth Phase of the internship is only for select few. In EIP 1, we selected top 10% of the interns and worked with them to implement a much complex project end to end. Phase Four involved one-on-one sessions, where a mentor will help you with your guided research, execution and either "Deployment Mode - 1000 fps" or "Research Mode - Paper Publication". Phase Four will start in January 2019. Details about this session will be shared on time.
- Upload Session 3 Assignments
- Upload Session 3 Notes
