AGB

Software Engineer | Python & Data Enthusiast | Cloud Infra | Materials Science PhD | Check my project:

[AGB @antgobar]

🚀 https://t.co/fTwUxJGytS is in its early days!
No email required, sign up & explore with demo data, available for download on the home page.
Your feedback is key, it will drive improvements!
Built to simplify #phdlife , #postdocs, & #research!
#SaaS #BuildingInPublic #WebDev

[AGB @antgobar]

now available at https://t.co/Fh03BaSaE7

[AGB @antgobar]

Current side project: LabMate https://t.co/OWc8qB8J2R an Electronic Lab Notebook of sorts, stack: #fastapi, #postgres, #htmx & Pico CSS. Deployed on Render’s free tier. Feel free to check it out and give some feedback via the contact form #sideproject #buildinpublic #prototype

[AGB @antgobar]

Couple of notes:
1. Renders free tier spins down the app after inactivity so your initial page load could take a minute
2. I am not a frontend dev 🤪

[AGB @antgobar]

@hjwp Yes possibly... e.g functions with numpy matrix operations & parse data & pandas. Errors in results aren’t immediately obvious so prepending ‘self’ where relevant makes it easy to check if data is being parsed correctly. Maybe I’ve answered myself and need smaller functions..

[AGB @antgobar]

@hjwp I guess being able to access (therefore test) class attributes quite easily compared to variables within stand-alone functions. Or maybe that's a sign that my functions are too long/complex and could do with breaking down into smaller units..

[AGB @antgobar]

@hjwp Great, thanks! All new concepts and frameworks to me so it’s been a great learning experience. I’ll appreciate TDD more by applying it to my own projects for sure. I’m emulating #66daysofdata btw (in case you’re wondering why someone is documenting their progress with your book!)

[AGB @antgobar]

@adamroyles Do you follow the TDD approach?

[AGB @antgobar]

@KOrfanakis I'd be very interested in seeing your code. Will you be uploading to kaggle?

[AGB @antgobar]

Day 66 of #66daysofdata Last day! So glad I got started on this @KenJee_DS! A true habbit now!
To finish off in style I analysed my twitter #Analytics 😅 more "#" != more engagement, but including media can help a bit. #Tensorflow is a good "#' for engagement #Statistics isn't😂

[AGB @antgobar]

Day 65 of #66daysofdata - getting back to the #udemy #DataScience to finish it. On the final section which focuses heavily on #TensorFlow 1.x, which is lowe level than TF2.x Should help me with version issues I've been having on my kaggle notebook...

[AGB @antgobar]

Day 64 of #66daysofdata - (messed up days again 😅)
Added #neuralnetwork model to Heart Disease #kaggle notebook. https://t.co/cSzp3i4u5o - with a lot of help from the #udemy course!
Several important hyperparameters can now be fixed (shown in plot title).

[AGB @antgobar]

Day 64 of #66daysofdata - learning a lot of pandas and seaborn. #python libraries for #kaggle project.
All very useful for #DataScience storytelling. Now to make #MachineLearning section visually appealing and user friendly enough for a notebook...
https://t.co/M1ktCNtgji

[AGB @antgobar]

Day 63 of #66daysofdata - looking at other Heart Disease UCI notebooks I realise that val accuracy of 86% achieved with a #neuralnetwork is not bad... Currently 'prettifying' my notebook with the seaborn #python library, though I need to learn more about subplots... #DataScience

[AGB @antgobar]

Day 61 of #66daysofdata putting kaggle project to one side while I brush up on p-values (important in medicine): the probability that an observed difference occured randomly, i.e. the significance of an observation.
https://t.co/zJKp24bfXP
#Statistics #DataScience

[AGB @antgobar]

Day 61 of #66daysofdata - not sure if this counts but today was a #python day... plotting a bunch of curves 😅
Happy #Halloween!!!🎃👻😱🧙

[AGB @antgobar]

Day 60! of #66daysofdata hyperparameter tuning on the #kaggle dataset with #python loops! Larger hidden layer size = same loss minima but less epochs required.
Results still not great, but fun paying around with functions. I need to look at smarter HP tuning... #MachineLearning

[AGB @antgobar]

Day 59 of #66daysofdata - a lot of troubleshooting #TensorFlow (version issues), but used a #NeuralNetwork model on on Heart Disease dataset. Results are... 😬 as predicted due to sample size! Batch size had a significant effect on the results. more HP tuning needed #DataScience

[AGB @antgobar]

Day 58 of #66daysofdata #Kaggle Heart Disease #DataScience project. Finalised preprocessing with sklearn #python library https://t.co/UpyTsABaVi
Next: create the #NeuralNetwork model. Only 303 observations, so a different #MachineLearning model maybe more suitable, we'll see...

[AGB @antgobar]

@timstevenson That’s right guys! Good effort 👍#DataScience

[AGB @antgobar]

Simple and useful things I've learnt:
1) #kaggle doesn't have the latest seaborn
Update in a cell:
!pip install seaborn --upgrade
Make sure "internet" is on (see pic)
2) plotting:
sns.kdeplot(df.a, df.b, fill=True, hue = df.c)
Compare a & b, with c as label
#python #DataScience