Main target: I will try to devote at least an hour to learn or implement something in ML/DL field
This repo is about maintaing my daily activities regarding the #60DaysOfUdacity Challenge
Start Date | End Date |
---|---|
27 June, 2019 | 27 August, 2019 |
- Make habit of working on Deep Learning
- Going deep into deep learning.
- Working on some projects and getting more experience on it.
- Getting the Nanodegree Scholarship.
Today's Progress:
- Worked on my thesis slides
- Organizing a meetup for the secure and private AI group
New thing(s) learned:
- Couldn't work any new thing as I am working on the slides.
Thoughts: Really need to finish the slides tomorrow.
Today's Progress:
- Finished my thesis slides
- Started on revising CNNs and GANs for my thesis viva
- Working on arranging some projects with BD group of SPAIC
New thing(s) learned:
- Couldn't work any new thing as I need to concentrate somewhere else. I will start working fully after 30th.
Thoughts: Will prepare for the slides tomorrow
Today's Progress:
- Defended my thesis
- Started learning from intro to pytorch again.
- Panning the meetup for 5th July
New thing(s) learned:
- Couldn't work any new thing as I was busy.
Thoughts: I will try to work much more from tommorrow.
Today's Progress:
- Read about Capsule net.
- worked on #ama_sessions as a moderator
- As I forgot some basics, I took the lesson 2 of intro 2 pytorch with DL again and took notes.
- Learning in depth on RNN and coding
New thing(s) learned:
- Learned about intuition of Capsule nets.
- Learned why perceptron algo works which I missed previously.
Thoughts: Will finish the RNN code tomorrow
Today's Progress:
- Started on Kaggle
- talking about forming groups
New thing(s) learned:
- How to submit on kaggle.
Thoughts: Couldn't do much because I was busy watching the match.
Today's Progress:
- Submitted for the first time on kaggle. Accuracy: 97% | Score: 0.9923 | Link: https://bit.ly/309E6xD | Will try to improve it more.
- Started http://fast.ai course - on lesson 1.
New thing(s) learned:
- How to submit on kaggle.
Thoughts: Couldn't do much because I was busy watching the match.
Today's Progress:
- Formed a study buddy group named Cognitive Learners
- Filled up the form for another one to form group with my fellow Bangladeshis.
- Right now working on the fastai course and taking notes.
New thing(s) learned:
- Working on communication skills
Thoughts: Love communicating with people...I get to learn so many new things from them.
Today's Progress:
- Attend first virtual meetup of #sg_cognitive-learners
- Revised the first chapter of The Algorithmic foundations of differential privacy which I have read earlier.
New thing(s) learned:
- Privacy
Thoughts: Will plan more to work on projects.
Today's Progress:
- Read this paper on hyperparameters: https://arxiv.org/abs/1803.09820 |
- Especially I read it to learn more about one cycle learning which helps to reduce training time for Deep learning
- Also read related articles about that.
New thing(s) learned:
- Learned about cycle learning
Thoughts: Will work more on it .
Today's Progress:
- Participated on the Women Who Code virtual meetup.
- Discussed with a teammate about our project steps.
- Started reading the related research paper for that
New thing(s) learned:
- Explored ideas about song recommendation
Thoughts: Really love attending these meetups.
Today's Progress:
- Read the first chapter of Grokking Deep Learning book by Andrew Trask to see how it is. And it seemed to be a good one | https://www.manning.com/books/grokking-deep-learning
- Finished week 1 of course 1 of Mathematics for Machine Learning Specialization| https://www.coursera.org/learn/linear-algebra-machine-learning.
- Revised some data structures from geeksforgeeks.
- Finding more ways to work on the project of #sg_cognitive-learners
- First meetup with our kaggle group from #sg_spaic-kagglers
- Discussed the upcoming group initiatives for #sg_cognitive-learners
- Working on the Be a Pytorch Beast initiative of #sg_novice-ai.
New thing(s) learned:
- Maths on ML
Thoughts: Hectic day it was!!! Loved the grokking book
Today's Progress:
- Revised the first four modules of Intro to DL with PyTorch as I need to help on PyTorch Beast Weak.
- Watched some Data analysis tutorials for our project of #sg_spaic-kagglers
New thing(s) learned:
- How to submit on kaggle.
Thoughts: Couldn't do much because I was busy watching the match.
Today's Progress:
- Decided the projects for #sg_cognitive-learners. We have three tracks to work on - advanced, medium, beginner. Interested people are welcome to look on.
- Worked on a personal project for Rice Leaf Disease classification. but lacking in dataset. ๐ Help is appreciated.
- Worked on math. Needed visualization for the vectors I learned from coursera specialization. So, I checked the Essence of linear algebra course of 3Blue1Brown again. That is the best series for Linear Algebra apart from MIT 18.06.
- Worked on some data preprocessing on Titanic dataset of Kaggle.
- Discussed initiatives of #sg_bangladeshi-schlrs with a friend
- Meetup with kaggle group from #sg_spaic-kagglers
- Reading and trying to understand a code on BERT that my friend provided ^_^
New thing(s) learned:
- Learned a lot of libraries regarding BERT which is similar to cyclegan on NLP
Thoughts: Realized taht I love working and it's better when I wake up early.
Today's Progress:
- Read "Generative Adversarial Nets" paper by Goodfellow.
- Finalized the projects for the group #sg_cognitive-learners I'll be working on a song recommendation project.
- Meetup with from #sg_caffeine_coders team
- Meetup with #sg_bangladeshi-schlrs team
- Meetup with #sg_spaic-kagglers team
- Teamed up for emotion detection project of #sg_caffeine_coders
- Finalized the project I'll be working on for #sg_bangladeshi-schlrs. After some searching, I decided to work on text summarization.
New thing(s) learned:
- Learned about a paper which solves simple math problems via neural network.
Thoughts: -
Today's Progress:
- FInished the GAN paper finally. But needs a looot of clarification!
- A few pages of Deep Learning book by Goodfellow
- Meetup with #sg_wrldwde-women-cdrs team.
New thing(s) learned:
- Learned that I have still a lot to learn on GANs.
Thoughts: I really want to finish this book.
Today's Progress:
- Revised math concepts
- Revised basic Deep learning concepts that I read this week.
- Mostly worked on the weekly Hackathon
New thing(s) learned:
- Hackathon is a new experience
Thoughts: I decided to revise the things I read the earlier day. So, didn't read anything new.
Today's Progress:
- Read about data preprocessing for Song recommendation project
- Mostly worked on the weekly Hackathon with the awesome hackathon team.
- worked a bit more on Dog breed classification.
New thing(s) learned:
- Learned more about transfer learning
Thoughts: Love working on teams ^_^
Today's Progress:
- Did a little data preprocessing for APTOS Blindness detection project
- Finally done with Hackathon. Thanks to my best team mates
- Meetup with the amazing ones @anne @Mohona @Droid @Mushrifah Hasan @Marwa for #sg_depressn-detectn
- worked on fastai course for a while.
- Worked a little on Rice Leaf disease classification.
- Another meetup with @Mohona @Munira Omar @Shubhendu Mishra @Marwa @Richaldo_Elias for #sg_spaic-kagglers
New thing(s) learned:
- Data preprocessing on Kaggle
Thoughts: Working on hackathon is exhausting
Today's Progress:
- Continued reading Deep Learning chapter 2 along with Essence of Linear Algebra videos to visualize better.
- Reading articles for emotion detection project.
- Tried out ways to improve digit recognizer for #sg_spaic-kagglers project
New thing(s) learned:
- Maths on ML
Thoughts: Still not good at transfer Learning
Today's Progress:
- Worked on rice leaf disease classification using Fastai and now I am in love with it.
- Almost done with Dog breed identification but stuck on a weird error.
- FInished training of Digit recognizer. Will finish the inference part tomorrow.
New thing(s) learned:
- Worked on my first FastAI project
Thoughts: Would like to work more on FastAI.
Today's Progress:
- Continued with fastai lesson-3.
- looked through some kaggle kernels to get better idea of the recognizing faces in the wild project.
New thing(s) learned:
- Learning a lot through the kernels
Thoughts: Someday i want to be a kaggle master!
Today's Progress:
- Revised Lesson 6 of style transfer and learned more in depth about Gram matrix
- Finally selected for emotion detection project for #sg_caffeine_coders
- Looked through some papers for the #sg_depressn-detectn with @anne @Mohona @Oudarjya Sen Sarma @Marwa @Mushrifah Hasan @Droid
- Started working on Hackathon with @Shahana @Joy Poddar @happycoder354 and still working ๐
- Attended a meetup with @Jess @Mohammad Diab @Shashi Gharti @Joyce @Elena Kutanov @Mateusz @Venkata Rathnam Muralidharan @Erika Yoon @Temitope Oladokun @Sergio Valderrama @Dammy @Suraiya Khan @Marwa @Mahedi @Sridhar Ethiraj @Joyce Obi @Sourav @Ayesha Manzur @erinSnPAI @Ahmed MAGDY EISSA @Oudarjya Sen Sarma @Muhammad Naufil @Sarah Majors @Sushil Ghimire @yasser albarbary @LauraT@Helena Barmer
New thing(s) learned:
- Loved the idea of Gram matrix Thoughts: Really want to work more on depression deteciton.
Today's Progress:
- Mostly worked on the hackathon.
- worked on tuning the hyperparameters, learned more about scheduler.
New thing(s) learned:
- Schedulers
Today's Progress:
- Worked till 11 am GMT(5 am my time) for the hackathon..still couldn't do much
- Revising fundamental concepts of deep learning.
- Attended a meetup with @Ashish Shrivastava @Mohona @Sushil Ghimire @sourav kumar @Aarthi Alagammai @Michael Sheinman @Muhammad Wajahat @Dammy @Govind in #sg_dlstarguys| Sorry everyone for being such a shaky host!
New thing(s) learned:
- Hosted a meetup for the first time
Today's Progress:
- Went through DL interview questions.
- Checked the Style transfer code again from lesson 6 or ud188.
New thing(s) learned:
- Interview questions on Deep learning.
Thoughts: I have my first interview tomorrow
Today's Progress:
- Restarted the lesson 3 of this course.
New thing(s) learned:
- I got the job :D
Thoughts: Will be out till 28th. So I'll just try my best to do at least something to continue my #60daysofudacity streak.
Today's Progress:
- Watched 3rd video of lesson 3 of this course.
- Got this great news that we were selected as the 2nd runner up of Auto-matic Hackathon.
New thing(s) learned:
- Today I had the most breathtaking day of my life. I with my friends trekked mountains with heavy rainfall and slippery ways. It's irrelevant but still I wanted to share it here!
- Got my first ever prize in hackathon.
Today's Progress:
- Been revising differential privacy. Since I'm still out of my place, I managed to start AFDP's second chapter.
Today's Progress:
- Worked on Malaria Project with @Rishi S Rao
- Looking for picture ideas and trying them on my style stransfer implementation for #sg_project-t-shirt with @Shubhendu Mishra and @Archit
- working on the ama faq doc
- Discussed about our meetup for #sg_caffeine_coders with @sourav kumar @Sabrina @Sumanyu Rosha
Today's Progress:
- Tried to increase the accuracy a bit more of Malaria Dataset.
- Rewatched the last few lessons of Essence of linear algebra series.
- Started learning Deep learning book of Goodfellow book. Couldn't read much today. But, I decided to read this book every day from tomorrow at least for 2 hours, inspired by @Stark
- Reading research papers for APTOS Blindness Detection
Today's Progress:
- Read up to section 2.11 of Deep learning book.
- Read papers for the APTOS Blindness detection. Learning more about Cohen's kappa coefficient that I stumbled onto one of the papers.
- Had an amazing meetup with @sourav kumar @Calincan Mircea Ioan @Oudarjya Sen Sarma @Aarthi Alagammai@Archit @Munira Omar @Mohona @Ashish Shrivastava@Govind @Sharim from #sg_dlstarguys
- Had another amazing meetup with @Jess @shahnoza@AbdElrhman @Alexander Villasoto @Munira Omar@David Ashraf @Sarah Majors @Hitoishi Das @Sourav@Nouran El Kassas @Shivam Raisharma @Mahedi from #sg_dl-goodfellows. Thanks @Shivam Raisharma for hosting it nicely
New thing(s) learned:
- Using SVD and PCA along with Deep learning.
Today's Progress:
- Learned about multitask learning.
- Read this one: https://arxiv.org/ftp/arxiv/papers/1604/1604.02878.pdf
- And this one: https://arxiv.org/pdf/1503.03832.pdf
New thing(s) learned:
- multitask learning
Today's Progress:
- Learned more about multi-task learning.
- Reading this paper: https://arxiv.org/pdf/1801.09414.pdf
- Attended the meetup of #sg_pytorch-robotics
Today's Progress:
- Worked on the APTOS Blindness Detection project. Reading research papers on it and implemented two of the papers according to it with @K.S.
- Discussing about depression detection project with @Mohona @Munira Omar @Droid for project showcase challenge
- Had a meetup about another project with @Stark for the same.
Today's Progress:
- Finally done reading all the expected papers for APTOS blindness detection and then discussed with @K.S. about them and next plan for the project.
- Started watching nlp tutorial from https://www.youtube.com/watch?v=8rXD5-xhemo&list=PLoROMvodv4rOhcuXMZkNm7j3fVwBBY42z&index=2&t=0s
- Read this Article on depression detection: https://time.com/1915/how-twitter-knows-when-youre-depressed/?source=post_page--------------------------
Today's Progress:
- Worked on Malaria Project with @Rishi S Rao
- Looking for picture ideas and trying them on my style stransfer implementation for #sg_project-t-shirt with @Shubhendu Mishra and @Archit
- working on the ama faq doc
- Discussed about our meetup for #sg_caffeine_coders with @sourav kumar @Sabrina @Sumanyu Rosha
Today's Progress:
- Completed the 2nd chapter of the Deep learning of Ian Goodfellow book and re-read the 1st chapter again.
- Read the following article and the articles attached mentioned there : https://medium.com/@jayeshbahire/introduction-to-word-vectors-ea1d4e4b84bf
- Finished the first lesson of cs224.
- Revised the notes of course 2 of DL specialization course.
- Attended the meetup of #sg_novice-ai with @Seeratpal K. Jaura @Stark @Hung @Archit @Oudarjya Sen Sarma @ayivima @Ingus Terbets @cibaca @Apoorva Patil @Alexander Villasoto @Agata [OR, USA] @Shudipto Trafder
Today's Progress: Had a really busy day...so couldn't do much at all.
- Revised about optimization, weight initialization, kl, js divergence.
- worked on an old project bug
Today's Progress:
Mostly it was a day with lots of discussions!
- Had a great meetup with @Munira Omar @Aniket Mitra@Ashish Shrivastava @Seeratpal K. Jaura@Anju Mercian @Mohona@Aarthi Alagammai @sourav kumar @Diganta for #sg_dlstarguys about math concepts.
- Discussed with @K.S. about our progress on APTOS Blindness detection. So far, we were able to work according to the papers and make our model runnable. Now, we need to work on accuracy.
- Had another meetup with @Mohona @Droid@Munira Omar @Aarthi Alagammai for our project of depression detection. So far we have scraped the dataset and our next target is to clean the data.
Today's Progress:
going through pretty busy days...so can't work much at all ๐
- Revised the course upto lesson 6
Today's Progress:
- Started Lesson 6 project finally.
- Data cleaning going on for the depression dataset.
- Had a meetup with @Aisha Khatun @Mahedi @Shudipto Trafder on GAN concepts for #sg_bangladeshi-schlrs
Thoughts:
- I really want to thank this slack community for making my birthday unforgettable ๐ I had an amazing birthday today for which a lot of credit goes to this community! I could never imagine so many people taking their time to wish me and thank you so much everyone and I had a blast!!
Today's Progress:
- Worked on promoting the math blogging competition for #sg_dlstarguys and #math_help
- Working on cleaning the depression data
Today's Progress:
- Had a meetup with @sourav kumar @Sabrina @Sumanyu Rosha about our project for #sg_caffeine_coders
- Had a meetup with @Pooja Vinod @Ingus Terbets@Shudipto Trafder @Oudarjya Sen Sarma @cibaca @Ivy@Ateniola Oluwatobi Victor @Ayushi Gupta @Stanislav Ladyzhenskiy @Mateusz@Sourav @Gogulaanand R @Sayed Maheen Basheer @Sushil Ghimire @susyjam @Helena Barmer @Frida @Abhishek Tandon @Stark @Ayesha Manzur, @Helena Barmer, @Jess, @Anju Mercian for #sg_study_jahm
- worked on APTOS Blindnes Detection and discussed with @K.S. about the same.
- Still cleaning depression dataset, it's never-ending
Today's Progress:
- Had a meetup for our depression detection project with @Mohona @Munira Omar @Aarthi Alagammai @Droid
- Worked on the APTOS Blindness detection project with @K.S. and managed to get more than 80% accuracy so far. Still working on the improvements.
- Worked on data preprocessing for the Emotion Recognition project with @Sabrina @Sumanyu Rosha @sourav kumar
- Started cleaning the depression data. Cleaning and reading the tweets giving me depression
Today's Progress:
- Almost finished cleaning the data...last 200 or some rows remaining.
- read this article: https://towardsdatascience.com/text-summarization-with-amazon-reviews-41801c2210b
Today's Progress:
- Finished the data cleaning for depression dataset finally ๐
- Read some part of this: https://ai.googleblog.com/2019/05/efficientnet-improving-accuracy-and.html
- Worked on APTOS Blindness detection for the submission...will submit it after this post
Thoughts: Couldn't submit :(
Today's Progress:
- Facing and solving the probles @K.S. and I are facing to submit the APTOS Blindness Detection kernel
- Changed the architecture of Malaria Detection for better accuracy
- Read this article: https://towardsdatascience.com/understanding-semantic-segmentation-with-unet-6be4f42d4b47
- Had a meetup with @SusanW and @Mohona for Depression Detection project
Today's Progress:
- Had an in detailed discussion meetup about our depression detection project with @Mohona @Munira Omar @SusanW @Marwa
- Discussed further plans about emotion recognition project.
- Worked little bit on APTOS blindness detection project..I couldn't find a way to submit the kernel while turning off the internet as I am using Transfer learning...so now I am trying a simple model and discussed about that with @K.S.
Today's Progress:
- Worked on APTOS blindness detection...still trying to improve the accuracy
- Discussed with @Sabrina about deployment of emotion recognition.
- Had a meetup with depression detection team - @Munira Omar @Aarthi Alagammai @Mohona @Marwa @SusanW
Today's Progress:
- Came out our previous depression dataset wasn't sufficient enough, so had to scrape again and now planning to use LIWC to extract the features.
Thoughts: Finally reached day 50!!!
Today's Progress:
- learned about siamese network
- Worked on depression detection Forgot to mention
- Attended #wmn_who_code meetup
- Learned about One shot learning and Zero shot learning
- Worked on aptos blindness detection.
Today's Progress:
- Corrected the dataset for Depression detection.
- Working with @Sabrina @Sumanyu Rosha @sourav kumar for emotion recognition project.
- Usual discussion with @K.S.
- Had a meetup with @SusanW @Mohona @Munira Omar @Aarthi Alagammai @Marwa
Today's Progress:
- Worked on the documentation of emotion recognition and aptos blindness detection project
- worked on solving a bug on emotion recognition project with @sourav kumar
- Had a meetup with the depression detection team
Thoughts: Couldn't submit :(
Today's Progress:
- Finished and documened Federated Blindness Detection project with amazing @K.S.: https://github.com/aksht94/UdacityOpenSource/tree/master/Federated%20Blindness%20Detection
- Finished and documented Federated Emotion Recogniton projet with @Sabrina@Sumanyu Rosha @sourav kumar: https://github.com/aksht94/UdacityOpenSource/tree/master/Federated%20Emotion%20Recogniton
- Finished Depression detection project with @SusanW @Marwa @Mohona @Aarthi Alagammai @Munira Omar @Droid: https://github.com/aksht94/UdacityOpenSource/tree/master/depression-detection
Thoughts:
- I want to thank @Soumia (you are an angel ^_^), @kkech and @Archit for the continuous help they provided me with..thank you ๐
- Celebrating and cheering my teammates @K.S. @Sabrina@Sumanyu Rosha @sourav kumar @SusanW @Marwa @Mohona @Aarthi Alagammai @Munira Omar @Droid It was an honour to work with you all
Today's Progress:
- Worked on improving the accuracy of APTOS blindness detection. Learned about using weights on loss functions...thanks to @K.S. โค๏ธ
- Had a discussion about pytorch Hackathon with @Shivam Raisharma @Halwai Aftab Hasan @cibaca @Hitoishi Das
Today's Progress:
- had a discussion with @Ashish Shrivastava about DL related applications
- Worked on APTOS Blindness detection for getting better accuracy...this project is my favorite one...love working on it ๐
- Looking for ideas for pytorch hackathon with @Halwai Aftab Hasan @cibaca @Shivam Raisharma @Hitoishi Das
- Encouraging @K.S. @Ashish Shrivastava @Mohona @Halwai Aftab Hasan @cibaca @Shivam Raisharma@Hitoishi Das
Today's Progress:
- Studied about a new optimizer aka Ranger. Going to try it on APTOS.
- Working on the Thank you Initiative for udacity-facebook. Would love to see everyone's participation. ^_^ Here: https://secureprivataischolar.slack.com/archives/CJSCX4WAZ/p1566660348242000
- Worked for the math blogging competiton.
- Watched this video that my friend @Stark made for easy deployment of PyTorch models using Katana I think it will help many from now on...^_^ https://youtu.be/VqeD8lRhl2o
- Attended the Deep Learning with PyTorch webinar as a presenter for the first time along with @K.S. @Pooja Vinod thank you guys for the support and managing so well..you did great ๐ Link to the slides: https://docs.google.com/presentation/d/1llOPlEZsMf6rPl7EJPutQvqedYyOE6sjAiCw66XxwRo/edit?usp=sharing
Thoughts: I am really thankful for this amazing community.
Today's Progress:
- Discussed about the ideas pytorch summer hackathon with @cibaca @Hitoishi Das @Shivam Raisharma @Halwai Aftab Hasan
- Worked on summarizing our projects for #project-showcase-challenge
Today's Progress:
- Went through many transfer learning articles to answer a question on #sg_novice-ai
- Attended the webinar of #sg_novice-ai | Thank you @Ingus Terbets @Alexander Villasoto for sharing your knowledge with us.
- Attended the meetup for PyTorch Summer Hackathon with @cibaca @Halwai Aftab Hasan @Hitoishi Das@Shivam Raisharma
Thoughts:
- I really really want to thank two of my friends @Stark and @Ashish Shrivastava for helping me to go on with this challenge...if it wasn't for you, I might not even complete this course! Hahah ๐
- Thanks @K.S. for the constant encouragement and support that you provided through working on APTOS Blindness Detection project...It would be really tough starting office if you didn't support