mejia

Speaking/building data and relationships for awesome people.
Discovering multiple four-leaf clovers in a vast field, perhaps suggesting a peculiarity.

[mejia @mejia_code]

Kolmogorov-Arnold Networks (KANs) use learnable spline-parametrized activation functions on edges, unlike MLP's (Multi-Layer Perceptrons) fixed (i.e., ReLU) on nodes. KANs offer interpretability vs. MLPs’ ‘black box’ and avoid forgetting.
https://t.co/HxsuIzgv1H

[Rerun @rerundotio]

A Rerun Viewer for the DROID Dataset!

DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset is a robot manipulation dataset by @SashaKhazatsky et al. with 76k demonstration trajectories or 350h of interaction data, collected across 564 scenes and 86 tasks.

[Physical Review Research @PhysRevResearch]

Unified framework for the entropy production and the stochastic interaction based on information geometry, Sosuke Ito, Masafumi Oizumi, and Shun-ichi Amari #statisticalphysics #entropy https://t.co/f1azIiXc1D

[peter! 🥷 @pwang_szn]

MY CABBAGES

[Hide @HideMatsu82]

Happy to share Gaussian Splatting SLAM

https://t.co/3VCrAeDfZ4

We show the first 3DGS-based Monocular RGB SLAM, the hardest SLAM setting.

Using 3D Gaussians as a unified representation, the method only requires RGB images - No need for SfM, depth sensor, or learned prior.

[Tai-Danae Bradley @math3ma]

I’m excited for this new article in the Notices of the AMS (@amermathsoc). In it, we describe several nice analogies between linear algebra and category theory and discuss connections to language and LLMs. Coauthored with John Terilla and @gastaldi_gianni.
https://t.co/rd3BB0MKOa

[Leonard Bruns @leonard_bruns]

Finding a textured mesh decomposition from a collection of images is a very challenging optimization problem. “Differentiable Block Worlds” by @t_monnier et al. shows impressive results using differentiable rendering. I visualized how the optimization works using @rerundotio.

[mejia @mejia_code]

Looking at GOES-16’s geostationary orbit vs. to other satellite orbits and distances from Earth. Its stationarity is due to its orbit matching Earth’s rotation, remaining fixed with respect to a point on the ground.
Great work @stuffinspace_!

[mejia @mejia_code]

Demo: normalized satellite data to generate false color images, aiding in the detection of contrails and the fight against global warming🛰️🌎. Happily, sharing an interim visualization (right: contrail) @nasa @NOAASatellites
https://t.co/5QGonqf0mc
https://t.co/dpxkNwMZ7R

[mejia @mejia_code]

Eye candy 59: A Goldberg polyhedron, denoted as GP(m,n), can be constructed by taking a ‘chess knight move’ from one pentagon to the next. i.e.A GP(1,0) is a dodecahedron with 20 vertices and 12 pentagonal faces (no hex faces). Thanks @beesandbombs for the inspiration!

[dave @beesandbombs]

:)

[Gabriel Peyré @gabrielpeyre]

The hairy ball theorem states that in odd dimension d, vector fields on the tangent plane of a (d-1)-sphere necessarily contain a singular point (where it vanishes). https://t.co/5UNiMnjI42

[mejia @mejia_code]

Mycoplasma mycoides JCVI-syn3B 🐐🦠
Engineering of a synthetic organism for specific functions while relying on their inherent 🧬adaptive capabilities @jaytlennon @IUBiology

[IU Biology @IUBiology]

Life finds a way. @jaytlennon and his research team found that a synthetically constructed minimal cell stripped of all but essential genes can evolve. https://t.co/c0Mr1NIxp9 @IUCollege @IUScienceNews @IUImpact

[mejia @mejia_code]

Optimizing path planning in robotics, enhancing stability using an improved 🐜 ant colony algorithm and high-order spline interpolation. Paving the way for safer, precise, and productive robotic systems." - Z. Xu, W. Wang, Y. Chi, K. Li, L. He: https://t.co/YoLU451CCU

[mejia @mejia_code]

Swin Transformer: New vision transformer.
Beats ViT/DeiT, ResNe(X)t models.
Results: Impressive on ImageNet-1K, COCO, ADE20K.
Unified architecture for vision, NLP.
arxiv: https://t.co/RPMH2NGVHX
repo: https://t.co/lsvGJ5gVGv
@OpenAtMicrosoft @MSFTResearch Asia

[mejia @mejia_code]

@NVIDIAAI superpower? 👇
i'm the captain, ingest my extensive codebase of aircraft maneuvers (>worth three stadium-sized libraries), and device a slingshot maneuver: set a course for a destination beyond the cislunar space. Engage!
@MetaAI @GoogleAI @neuralink? NVIDIA🇹🇼

[mejia @mejia_code]

neural graph db

1️⃣assume data gaps
2️⃣infers link between 2 nodes-edge
2️⃣creates a geo📐 proximate-embedding

∴
☑️re-train 💰
☑️updatability to unseen data
☑️real⏱️ gap solver, compiler?

https://t.co/0kCeWYGcWb @zhu_zhaocheng @michaelcochez @michael_galkin @ren_hongyu

[Andrei Kashcha @anvaka]

Map of @github have arrived. Hope you enjoy it:

https://t.co/WzoBuevSxk

[mejia @mejia_code]

need: training datasets that represent the full output stream generated by LangChain's chain-of-thought

@reedbndr @langchain let's sim some datasets?https://t.co/Y2E0V2EKel

ie. D&D dataset: https://t.co/ZLvBPhu5fx

[Reed Bender @reedbndr]

I’m trying to fine-tune an open-source LLM to function seamlessly as a chat agent chain in @langchain …

Looking for training datasets that represent the full output stream generated by Langchain’s chain-of-thought.

Anybody have suggestions? 🧐

[Ryohei Sasaki@engineer @rsasaki0109]

MPCC
Simulation environments in C++ and Matlab of the Model Predictive Contouring Controller (MPCC) for Autonomous Racing developed by the Automatic Control Lab (IfA) at ETH Zurich
https://t.co/JPNHtDvdly

[mejia @mejia_code]

TestN PrivateGTP

Arch: LLMs & embeds interpret queries for local generations

Data: Split, embed, & store locally w/ LlamaCppEmbeddings

Query: Retrieve w/ LLMs, Chroma object & RetrievalQA

🧠@ivanmartit @nomic_ai @langchain @ggerganov @trychroma @MetaAI @abetlen