Follow me on Twitter @Incept_shawn https://twitter.com/Incept_shawn for daily updates!
Today's Progress : Completed week 1 of the ML course on Coursera by Prof. Andrew Ng.
Today's Progress : Brushed up on my linear algebra. Did a little of week 2 of the course.
Today's Progress : Completed week 2 of the ML course on Coursera by Prof. Andrew Ng.
Today's Progress : Worked on (but not completed) week 3 of the ML course on Coursera by Prof. Andrew Ng.
Today's Progress : Decided on a definite curricullum to follow in my quest to learn ML and DL. (Includes a list of content from various sources like Coursera, Siraj's videos and 3blue1brown) Watched 3Blue1Brown's entire playlist on Neural Networks for today.
Thoughts : Right now, I have a fairly good understanding of the working of a neural network. Really exicted! Siraj anounced the challenge today and I decided to participate. I hope this daily commitment pushes me to work consistently.
Today's Progress : Had a look at the code behind a simple neural network from Siraj's video "Neural Networks in 4 minutes". Looked properly at each line of code and made sure I understood what everything was doing by connecting it with the 3Blue1Brown playlist and googling stuff I didn't know. Once I understood everything, I wrote the code in a Jupyter notebook myself and obtained the results
Thoughts : Today was a really busy day. I had to collect books in the morning and I had my cousin's birthday in the evening. And after that, I was making memes on the 'Sweden vs England" World Cup match. :P Had to salvage time in the afternoon to finish this. Hopefully tomorrow will be more productive!
Today's Progress : Completed week 3 of the ML course on Coursera by Prof. Andrew Ng. Also worked on (but not completed) week 4 of the course.
Thoughts : Today was not really better than yesterday. I begin my second year of college tomorrow and I had a few things to take care of which took up way more time than I expected. Anyways, I don't feel like I'm behind schedule but I should be putting in more effort everyday to make the most of this challenge. Let's see if I can have more written in "Today's Progress" for tomorrow than in "Thoughts". :P
Today's Progress : Began watching the videos of the Convolutional Neural Networks course on Coursera by Prof. Andrew Ng.
Thoughts : Today was not the best day for ML. Began the new year at college today. I can see how the next few months are going to really test my dedication to this challenge. I only managed to start with learning CNNs and I have a rough idea on what they are.
Today's Progress : Finished watching Week 1 of the videos of the Convolutional Neural Networks course on Coursera by Prof. Andrew Ng.
Thoughts : Returned home late from college with a bad headache but it got better as the evening progressed. Still managed to put in my daily effort for the challenge. This reassures me that I can stick with it because (barring unforseen circumstances) today was as bad as my day can get.
Also, CNNs sure are something!
Today's Progress : Rewatched the content on Neural networks and CNNs to strengthen my understanding.
Thoughts : Decided to slow down a bit and make sure I have my basics strong. The more I revise this, the more I'm beginning to appreciate the concepts.
Today's Progress : Strengthened my linear algebra understanding by digging further into 3Blue1Brown's videos. Also worked on the task of week 4 of the ML course by Prof. Andrew Ng.
Thoughts : Day 2 of revision. One reason for revision is that I'm too drained to learn new stuff after coming home from 8 hours of college. I plan on compensating with weekends.
Today's Progress : Completed week 4 of ML course by Prof. Andrew Ng. Also successfully implemented basic linear regression with gradient descent on my own based on my understanding without use of ML libraries.
Thoughts : Struggled quite a bit with the coding part even though my concepts were clear. Need to revise Python? Will upload the code soon and regularly work on improving it. UPDATE! Uploaded the code.
Link: https://github.com/recurshawn/Basic-Linear-Regression
Today's Progress : Went to the Goa IT day celebrations today and attended a panel discussion on using ML and AI to solve real world problems. Also did a little of week 5 of Prof Andrew Ng's ML course.
Thoughts : The event was good. It's a 2 day thing so I'll be going again tomorrow. It was fun listening to established people speak their opinions on tech topics. My state, Goa, is proposing a new IT policy that will boost the IT environment and startup culture here. Super excited for that!
Today's Progress : Went to the Goa IT day celebrations for the second day. Revised regression and classification and coded examples too.
Thoughts : Day 2 of the event involved launchig of the actual policy. They want to make Goa an IT hub and a leader in AI/ML amongst other technologies. Also, I networked a good amount! Really good day. :D
Today's Progress : Played around with my linear regression code. Tried modifying it to become quadratic regression but couldn't succeed. Will attempt it again tomorrow. Also worked on week 5 of the ML course by Prof Andrew Ng. Uploaded the code to linear regression programme on GitHub. Check it here:
Link: https://github.com/recurshawn/Basic-Linear-Regression
Thoughts : Not able to modify linear regression quadratic regression. I avoided looking up the code online because I want to figure it out on my own. I did, however, look up stuff related to Python the language I'm using. I think my logic is correct but somewhere there is a limitation in my code. Perhaps it's too taxing on my slow device. I'll try writing something more efficient tomorrow.
Today's Progress : Got my quadratic regression code to work. Learnt about tinkering with the learning rate and a new type of cost function. Solidified my understanding of gradient descent.
Thoughts : Finally got it to work! On the suggest of a friend, I used the absolute difference as a cost function instead of squared difference and it helped. Of course, I also had to mess with the learning rate to make sure the descent converges. I actually used an incomplete derivative in the descent initially but surprisingly got the desired output. My theory is that the magnitude of the derivative does not matter so much as the direction it provides to facilitate gradient descent. In this case, the incomplete derivative still provided the correct sign. I was so sleepy yesterday, I couldn't even type this log. In fact, I'm very tired today too even though I didn't do as much as I did yesterday.
Today's Progress : Played with the code of my quadratic regression to try to see if I could get it to work with the squared difference cost function. Watched a little of week 5 of the ML course by Andrew Ng.
Thoughts : Couldn't succeed with the squared difference cost function. I may come back to it later. It's time to move on. I need to pick up pace. Today was another slow day, college is really affecting my productivity. I need to set weekly ambitious deadlines for myself. I'll upload the code for my quadratic regression later this week.
Today's Progress : Completed watching week 5 of the ML course by Andrew Ng.
Thoughts : The end days of the week are so tiring, I didn't feel like doing anything after coming back home. I need to shift some of my ML work to mornings when I have much more brain power. I am practially sleep typing thiiii..s lol.
Today's Progress : Explored the normal equation in depth and successfully implemented it in Octave.
Thoughts : It was going to be another unproductive day but I didn't let it come to that. The normal equation is so elegant! I never really appreciated it before. It's amazing how one simple equation can replace gradient descent (most of the time).
1/5th or 20% of the challenge done! I really hope I can do something amazing before day 100!
Today's Progress : Attempted building a basic classifier that distinguishes between coconut palms and mango trees based on fruit size and leaf length.
Thoughts : This was pretty challenging. But very cool. Trying these things out on my own is really helping me grasp everything better.
Today's Progress : I attended my first GDG (Google Developers Group) meetup and learnt how to build Actions for the Google Assistant.
Thoughts : Loved the meetup! It's so much fun and it's really refreshing to meet with like minded people. I'm now really curious to explore NLP further. Oh, and I'm defintely going to be regular with these meetups.
Today's Progress : Began coding a basic neural network (again)
Thoughts : Not very productive day (blame college). I began writing a basic neural network based on my own understanding.
Today's Progress : Finished working on that neural network and began playing around with it to mimic logic gates.
Thoughts : Wasn't successful and I couldn't pursue it further. Ironically, because I had to complete my Logic Design file. XD
Today's Progress : Watched a little of week 2 of the CNN course by Prof. Andrew Ng.
Thoughts : Didn't do the neural network thing from yesterday because today was another busy day. Fingers crossed for tomorrow.
I haven't updated the log daily from July 26 till August 17 due to lack of time and lots of other stuff I had to do besides the challenge. Therefore, most of the entries in this time period will be brief as I used my tweets as a reference to fill them in. I will try to be more regular with future updates.
Today's Progress : Was not able to find the error in my code, so I played around with Trask's code to simulate logic gates and also a simple mango/coconut tree classifier. Thoughts : The error took up too much time and I was still unsuccessful at detecting it. :(
Link: Here is the code I modified (courtesy: Trask) : https://iamtrask.github.io/2015/07/12/basic-python-network/
Today's Progress : Watched a little of week 2 of the CNN course by Professor Andrew Ng and also brushed up on my understanding of probability particularly conditional probability and Bayes theorem. Thoughts : Didn't waste time with error in my NN code.
Today's Progress : Solved the issue with my own NN code. Error was in transposing an array.
Thoughts : FINALLY DID IT!! Apparently, you can't directly transpose a 1D array in Numpy. I had to resolve that by directly initialising as a column matrix. Whew. This was a trying week!
Today's Progress : Started with week 6 of the ML course by Prof. Andrew Ng.
Thoughts : Started thinking about what project I want to take up in the later half of my challenge.
Today's Progress : Continued with week 6 of the ML course by Prof. Andrew Ng.
Thoughts : Was occupied with other work so couldn't do much. Ugh, why does college have to come in the way of everything?!
Today's Progress : I started bingeing on YouTube videos of giant_neural_network.
Thoughts : I had stumbled on his channel last year and I had implemented basic regression on my own only based on the concept he explained without having a look at anyone's code. He makes it sound so intuitive! You have to check him out. Link : https://www.youtube.com/playlist?list=PLxt59R_fWVzT9bDxA76AHm3ig0Gg9S3So
Today's Progress : Watched more of giant_neural_network videos and also did a little of week 6 of the ML course by Prof. Andrew Ng.
Today's Progress : Explored Colaboratory and watch Siraj Raval's (@||Source||) video on AWS. Also, cleaned by desktop and installed Fedora on it.
Thoughts : Unfortunately, I won't be able to use TensorFlow on my desktop as I don't have the specs hence I will be using Colaboratory for when I get started with TensorFlow.
Today's Progress : Began honing my CS skills in recursion, sorting and data structures. Also finished watching the Neural Network playlist on giant_neural_network.
Thoughts : That playlist is soo good, I really recommend checking it out! Also, be sure to subscribe to his channel. He defintely deserves way more than 13K subscribers! Link : https://www.youtube.com/playlist?list=PLxt59R_fWVzT9bDxA76AHm3ig0Gg9S3So
Today's Progress : Worked on week 6 of the ML course by Prof. Andrew Ng.
Thoughts : It's been a really slow week for ML :/
Today's Progress : Attended the International Blockchain Congress 2018 held at Panaji, Goa.
Thoughts : I was exposed to the wonderful world of Blockchain. It's amazing how simple ideas can be so impactful and useful. Blockchain is definitely going to be at the back of my mind while I progress through this ML challenge.
Today's Progress : Worked on recursion and understood it. Understood and coded the solution for Towers of Hanoi problem in C using recursion.
Thoughts : Recursion is a little complicated to implement but it gets intuitive with practice. Mind blowing stuff :P
Today's Progress : Coded Tic-tac-toe the game in C++
Thoughts : Didn't realise it was so simple to do. Managed to do it in under an hour. I could upgrade it by making an AI opponent as a player but the game is so simple, I could just use "if-else" statements for the same effect! XD
Today's Progress : Began learning and implementing different sorting techniques like selection, bubble and insertion sort. Also caught up with latest news in AI.
Thoughts : Am I cheating in this challenge by making college work and challenge work the same thing? :P It gets really tiring balancing both.
Today's Progress : Reviewed CNNs. Played around with classes and objects in C++.
Thoughts : Ugh, I had to spend so much time today completing a file for college.
Today's Progress : Learnt and understood eigenvalues and eigenvectors in linear algebra.
Thoughts : Linear algebra is not taught in the most fun way in college but I can't recommend the 3Blue1Brown playlist on linear algebra enough! Good stuff! Link : https://www.youtube.com/playlist?list=PLZHQObOWTQDPD3MizzM2xVFitgF8hE_ab
Today's Progress : Got access to a TensorFlow on Google Cloud Platform specialisation.
Thoughts : I got access to a TensorFlow on Google Cloud Platform specialisation thanks to GDG Goa. I will be working on completing it during the next few weeks.
Today's Progress : Attended the GDG Goa event on "The Art of Machine Learning" and learnt on how to go about building products with ML, building chatbots for businesses etc. Also got introduced to Computer Vision.
Thoughts : A really good event! Learnt a lot of new stuff. I was also informed of the MLCC Study Jam by Google. I am definitely signing up for that.