Conor O'Sullivan
@conorosullyds.bsky.social
📤 464
📥 955
📝 160
PhD in progress - XAI and ML for coastal monitoring 🌊 My content: linktr.ee/conorosullyds
pinned post!
Explainable AI (XAI) is about illuminating black-box machine learning models and explaining them in a way that we can understand. I'm fascinated with this topic. But learning about it was a struggle as there's not much educational content out there. So I made a course. And it's free!
loading . . .
Explainable AI for Computer Vision: Free Python Course - A Data Odyssey
A free course for the theory and Python code for XAI methods including Grad-CAM, SHAP, Occlusion, DeepLIFT, Integrated Gradients and Deconvolution.
https://adataodyssey.com/xai-for-cv/
12 days ago
0
1
0
It's unlikely that LLMs have consciousness, but just in case, I'm treating mine to some dumpling soup.
about 1 hour ago
0
0
0
I've made some updates to my explainable AI course. All the gradient-based methods now have articles. Videos coming soon :)
adataodyssey.com/xai-for-cv/
loading . . .
Explainable AI for Computer Vision: Free Python Course - A Data Odyssey
A free Explainable AI (XAI) course for the theory and Python code: Grad-CAM, SHAP, Occlusion, DeepLIFT and Integrated Gradients and more!
https://adataodyssey.com/xai-for-cv/
about 4 hours ago
0
0
0
Cover image from my article on DeepLIFT :)
about 4 hours ago
0
0
0
One of the biggest problems with saliency maps is noisy gradients. My latest YT video explores how you can remove that noise by adding noise. See how SmoothGrad can be used in conjunction with other explainable AI methods 👇
youtu.be/DoG6KtbtFvg
loading . . .
Cleaner Saliency Maps with SmoothGrad | XAI for Computer Vision
YouTube video by A Data Odyssey
https://youtu.be/DoG6KtbtFvg
1 day ago
0
0
0
Saying please and thank you to an LLM is the modern version of Pascal's wager
2 days ago
0
1
1
Been doing a bit of snowboarding in Turkey this past week
3 days ago
0
0
0
Challenge for those who are very confident that there isn’t a Flying Spaghetti Monster in the sky that controls the universe: Explain how the universe works.
add a skeleton here at some point
3 days ago
0
0
0
I thought the ML community on bluesky was dead before following some of this advice 👇🏼Article by
@nsaphra.bsky.social
nsaphra.net/post/bsky/
loading . . .
The AI Researcher's Guide to a Non-Boring Bluesky Feed | Naomi Saphra
How to migrate to bsky without a boring feed.
https://nsaphra.net/post/bsky/
3 days ago
0
22
3
If the Americans are going to turn towards authoritarianism they should learn from the Spanish and get a guy who is really into trains
3 days ago
0
0
0
I’d settle for being a Jack of 1, maybe 2 trades
3 days ago
0
0
0
loading . . .
Vanilla Gradients - A Data Odyssey
Explore vanilla gradients for explainable AI — how they work, their limitations, and a practical Python implementation using VGG16.
https://adataodyssey.com/vanilla-gradients/
6 days ago
0
0
0
loading . . .
Input X Gradients Explained: Why This XAI Method Can Mislead You
YouTube video by A Data Odyssey
https://youtu.be/20PkvlbyTNQ
8 days ago
0
0
0
An animation for an upcoming video on DeepLIFT. This is by far the most time-consuming one I've made.
loading . . .
12 days ago
0
0
0
Three different definitions of Explainable AI
12 days ago
0
0
0
Explainable AI (XAI) is about illuminating black-box machine learning models and explaining them in a way that we can understand. I'm fascinated with this topic. But learning about it was a struggle as there's not much educational content out there. So I made a course. And it's free!
loading . . .
Explainable AI for Computer Vision: Free Python Course - A Data Odyssey
A free course for the theory and Python code for XAI methods including Grad-CAM, SHAP, Occlusion, DeepLIFT, Integrated Gradients and Deconvolution.
https://adataodyssey.com/xai-for-cv/
12 days ago
0
1
0
reposted by
Conor O'Sullivan
Theo Sanderson
30 days ago
I made a map of 3.4 million Bluesky users - see if you can find yourself!
bluesky-map.theo.io
I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail
loading . . .
Bluesky Map
Interactive map of 3.4 million Bluesky users, visualised by their follower pattern.
http://bluesky-map.theo.io/
659
7205
6741
I think this is more of a warning about being a matplotlib contributor than anything else
theshamblog.com/an-ai-agent-...
loading . . .
An AI Agent Published a Hit Piece on Me
Summary: An AI agent of unknown ownership autonomously wrote and published a personalized hit piece about me after I rejected its code, attempting to damage my reputation and shame me into acceptin…
https://theshamblog.com/an-ai-agent-published-a-hit-piece-on-me/
13 days ago
0
0
0
How a CNN makes predictions. Earlier layers may extract certain features like edges and textures from the input. These are then combined in deeper layers to create features representing specific objects, like pieces of sushi.
14 days ago
0
0
0
I yearn for simpler times
add a skeleton here at some point
14 days ago
0
0
0
The real tragedy is that AI has killed the novelty of those Photoshop edits where they combine two animals
14 days ago
0
0
1
Finally going to try YouTube's A/B testing. Which Thumbnail do you think will do better?
15 days ago
0
0
0
New videos coming soon ✌️
youtu.be/sz964mAVcAE
loading . . .
Channel Update | Free Course, Future Videos & PhD Progress
YouTube video by A Data Odyssey
https://youtu.be/sz964mAVcAE
15 days ago
0
0
0
A little animation from my upcoming video on LIME
loading . . .
15 days ago
0
0
0
I'm definitely not one to overhype AI, but saying it is useless is just wrong. My dad is 60 years old, and he can stop talking about how helpful it is to his painting manufacturing business. He's planning to expand into a bunch of products with its help.
15 days ago
0
0
0
Done editing! I've moved over to a new mic, so hopefully these come out well.
19 days ago
0
1
0
Loss landscape gif created using Python code from Claude. LLMs really are an amazing storytelling tool.
loading . . .
20 days ago
0
1
0
A little animation (or catimation, rather) from an upcoming video on smoothgrad.
loading . . .
20 days ago
0
1
0
SmoothGrad: adding noise to remove noise Working on the gradient-based section of my course. Going to turn this image into a little animation.
21 days ago
0
0
0
Actually I was wrong. This problem has more to do with the linearity assumption.
add a skeleton here at some point
26 days ago
0
0
0
I've been digging into it a bit, and I think the Captum implementation of KernelSHAP is super misleading. It uses a baseline. Although this is practical for image data, it means we are really applying Baseline SHAP (BShap).
captum.ai/api/kernel_s...
loading . . .
Captum · Model Interpretability for PyTorch
Model Interpretability for PyTorch
https://captum.ai/api/kernel_shap.html
26 days ago
0
0
0
Glass Beams make the best music to work/focus to
music.youtube.com/watch?v=v_mp...
loading . . .
Black Sand
YouTube video by Glass Beams - Topic
https://music.youtube.com/watch?v=v_mpRUw_JDk&si=cmLQDeNRtI3om8cy
26 days ago
0
0
0
I've been reading a lot of Roddy Doyle lately
26 days ago
0
0
0
One of the major limitations of applying KernelSHAP to image data. Features (i.e. clusters of pixels) tend to be highly correlated. This means we can only get reasonable results when we use large superpixels that are approximately independent.
27 days ago
0
0
1
One of my favourite uses of LLMs:
27 days ago
0
0
0
Machine learning models are complex functions. The idea behind LIME is that this complexity falls away if we zoom into the feature space in the area around an instance. The function is much simpler or even linear.
28 days ago
0
2
0
A w.i.p. figure for my LIME article. I want to show how, although the original model uses images, we end up training a surrogate model on tabular features.
28 days ago
0
0
0
Working on the Integrated Gradients section of my Explainable AI course. One of the more computationally expensive gradient-based approaches. Thankfully, it tends to converge with around 50 backwards passes.
29 days ago
0
1
0
The steps behind Integrated Gradients We calculate the gradients for every image along a path from the baseline to the input image. Talking the average of all those gradients and multiplying by the difference between input and baseline gives us our final saliency map.
about 1 month ago
0
0
0
I have once again made the mistake of expressing my opinion on the internet
about 1 month ago
0
0
0
Now that the paper is out, I'm excited to share some of the results over the next week :D
add a skeleton here at some point
about 1 month ago
0
1
0
My latest paper has been published in JSTARS. Shows how it is possible to detect the coastal vegetation line with Sentinel-2 data with far more precision.
ieeexplore.ieee.org/document/113...
loading . . .
Detecting Sub-pixel Changes in the Coastal Vegetation Line with Sentinel-2 Imagery
Monitoring coastal erosion often requires detecting changes in the vegetation line (VL). Traditional approaches rely on aerial photography and manual annotation, restricting analysis to a limited numb...
https://ieeexplore.ieee.org/document/11372985?denied=
about 1 month ago
0
0
1
Working on the DeepLIFT section of my course
about 1 month ago
0
0
0
Working on the LIME section of my XAI course :)
about 1 month ago
0
0
0
I am once again making YouTube videos, and I am once again taking a bunch of cheesy headshots.
about 1 month ago
0
0
0
If you make any type of talking head content, Gling is a game-changer. Now can be get some of these features in final cut?
about 1 month ago
0
1
0
New setup for YouTube videos. Excuse the janky softbox.
about 1 month ago
0
0
0
I'm finally working on my course again. Will have some new videos out soon!
adataodyssey.com/xai-for-cv/
loading . . .
Explainable AI for Computer Vision: Free Python Course - A Data Odyssey
A free course for the theory and Python code for XAI methods including Grad-CAM, SHAP, Occlusion, DeepLIFT, Integrated Gradients and Deconvolution.
https://adataodyssey.com/xai-for-cv/
about 2 months ago
0
0
0
Input x Gradients is a simple XAI, but don't be fooled by the visually pleasing results. The saliency maps can be biased by dark and light pixels.
about 2 months ago
0
0
0
The limitations of raw/vanilla gradients when it comes to explaining model predictions. This is why we need methods like DeepLIFT or Integrated Gradients.
about 2 months ago
0
0
0
A summary of axioms in XAI research and some popular axiom-based methods. Trying to show which axioms the methods satisfy. Which visualisation do you think does this best?
about 2 months ago
0
0
0
Load more
feeds!
log in
