So, I just completed the DL Specialization and even though Dr. Andrew said to be proud, I cannot. The reasons are as follows:
I do not remember most of the things which I learned and it feels like that I'll have to watch all the videos again.
The labs were fully guided and it has instructions which helped me to apply the knowledge that I gained through videos but I cannot implement the knowledge if I work on a problem completely different from the ones in the labs.
People are making so many things and even though I try a lot, I end up failing. It feels like that all the knowledge that I gained is just wasted. All my efforts gone to waste.
How to deal with this situaton and what can I do so that I can also land a job and contribute to improve this world by using AI?
Hoping to get supportive answers from the experts.
If you look at the course listings, deeplearning.AI estimates something like 25 ~ 30 hours per class. So after the Specialization you have spent, what 100 hours? 200? No one should self doubt over failure to master such a broad and complex subject in so little time. I suggest reinforcing your learning by rewatching the videos (you likely pick up on different nuances each time) and revisiting the programming exercises, especially places where you didn’t deeply understand either the learning objectives or the solution. Find some open data sets and implement some simple models to learn from them, then make your models more sophisticated, or the data sets larger, forcing yourself to deal with more complexities such as datasets that can’t fit into memory. Go back and read the foundational papers you were introduced to, and try to ensure you understand what problem they were trying to solve, how their approach moved the state of the art forward, and what remained ( and remains) to be done. Finally, recognize that learning is a journey. No one (other than maybe a handful like Prof Ng, Yann LeCun, or Chris Manning) really know everything about everything ML. Be patient.
In learning anything new, I find useful to apply the the field of my interest. For example, learning a foreign language as Russian, you can make much faster and deeper progress if you do an activity you’d normally do in your language in the language you are learning. Maybe you’d play Russian computer games or would play with Russian speaking friends. In the case of ML courses, perhaps practicing with a dataset that is related to your hobby/fields of interest. Anyway, you got this!
Sorry you’re feeling discouraged.
I have to admit, I feel a little the same way so I can relate.
I feel like I got a good understanding of everything we did in the class. I did all the optional labs, and read several of the source papers, and lots of browser tabs were opened and read along the way. Yet it nags me that I feel like if I were to sit down with a blank VS Code window and try to do anything, I couldn’t actually do anything.
But I like what ai_curious said. You know, they say 10,000 hours to master anything so I guess we’re probably about 1-2% towards that mastery.
To resolve this myself personally, I still plan to take a few more classes and I feel like that will help somewhat, but then after I’ve gotten all the specializations out of the way, nothing can boost your confidence like actually doing. Actually coding up your own toy projects from scratch (it’s OK to start with a tutorial handy on a nearby screen) I’m sure will cement these concepts in our minds and give us practice – where the rubber hits the road. I also plan to try some Kaggle contests, not with any goals to win but to get my hands dirty and get more comfortable with the tasks. You might try the same.
Hi there, I would say I was the same the first time I read over the material. It is a lot to digest in so little time and practice. One thing that helps me is to question what I learned and what I did not learn, i.e. learn to differentiate other problems. In this course you are learning classification problems, that is take images, and label them accordingly. But there are many other questions that NN or Deep Learning can answer. In some of them, you can translate some of this knowledge, but you will still require to do some investigation to enhance it. As people suggest, set yourself a goal with something you like and start asking what other tools you need. Good luck.
Hello @annoyingCode - first of all: Congratulations! you just completed the DL Specialization!
That’s impressive - you’re in a relatively small group of people who have succeeded. Are there people who are further along this path? Sure! But there’s also a nice phrase from martial arts masters:
