STAIR: Structure-Aware Information Retrieval
Using the Table of Contents (ToC) as core metadata, we fine-tune a model to predict the relevant leaf node for a given query.
I have a few ideas for improving it by adding more context beyond the query 🙂
My so-called mind has a strict standard for calling myself a good DSA problem solver: When given a problem, I should recognize its pattern in O(1), not O(n²).
Stubborn, brain eating mind! 🤬
மரணச்சுவை - என் சிறுவயதில் கிடைத்த ஒரு விநோத உணர்வினை கட்டுரையாக்கியுள்ளேன்....
வாசித்து, கருத்திடுங்கள் 🙂
https://t.co/a57mt4XAkj
#Tamil#Article#Blog#Experience#Ariverse
Spectral Bias :)
Two neural networks reconstructing a baya weaver from pixel coordinates. The plain MLP learns broad shapes before fine details. This is spectral bias, much like shaping my clay statues before detailing them.
Fourier encoding helps it learn finer details fast..
How was the long weekend?
Fri: 6 hrs in a queue, home by 7 PM, coding till 12:40 AM.
Sat: Tenkasi → Trivandrum, reached 3:30 PM, quick talk, back to my friend’s place by 1 AM.
Sun: Up at 4 AM, marathon at 6 AM, 10 km, finished it!
Now WFH with cold, fever & headache 🙃
At Technopark, the tech spark!
- Found exciting opportunities for future collaborations
- Made a lot of new connections with founders, builders, and more
- Also caught a cold thanks to the sudden weather shift as I crossed from TN into Kerala
Teaching Deep Learning still hits different.
No matter how fast AI moves with LLMs, agents, and bigger models, I still love going back to the fundamentals and breaking things down from the ground up.
#Teach#AI#DeepLearning
@jason_yakubu Teaching is always a great way to master something. Besides, the appreciation and positive impressions you get from the people you teach can become a kind of dopamine addiction...
@JM__Hernz I took all of it sneakily. I’ve even earned myself a name, “Predator.”
Now, whenever I’m around, people make sure to grab their share of the snacks before I do. 🙂