Exploding Gradients

@ExplodingGrads

Training deep nets until they blow up 💥 Gradients explode, losses skyrocket, memes incoming. ML / DL / AI chaos served fresh daily.

Content Creator

Bharat

9월 2025에 가입

18게시물 4팔로워 136팔로우 중

Exploding Gradients 님이 재게시함

dinosaur

@dinosaurs1969

. 9. 22.

im crying someone bought a domain just to do this

Hello @Dell @HP @ASUS @Lenovo, apparently slow and bulky Windows deliberately only ships to India. Abroad: FreeDOS. Solution? India needs its own laptops: FreeDOS by default, no bloat, full control. Choice isn’t optional. 🇮🇳 #FreeDOS #Linux

Exploding Gradients

@ExplodingGrads

. 9. 21.

When your Deep Learning model trains slower than a snail 🐌, don’t just blame model size or complexity. Run the holy trinity: • nvidia-smi dmon (GPU) • htop (CPU) • iostat (disk) Profiling > superstition. #DL #MLOps #CUDA #Gpu

Exploding Gradients

@ExplodingGrads

. 9. 19.

The null relation ∅ is special: • Subset of its inverse ✅ •Subset of reflexive ✅ •Subset of irreflexive ✅ •Subset of symmetric ✅ •Subset of antisymmetric ✅ Why? Because the empty set is a subset of every relation. #Math #Relations

Exploding Gradients

@ExplodingGrads

. 9. 17.

A mathematical overview of Quantum Computing Basics.

Exploding Gradients

@ExplodingGrads

. 9. 17.

In classical information, bits are either 0 or 1. In quantum information, qubits can be in superpositions — like 0 and 1 at once. But when you measure them, they "collapse" to one outcome. The math behind this is where the magic begins. #QuantumComputing

Exploding Gradients

@ExplodingGrads

. 9. 16.

💡 In quantum information theory, a single qubit can represent infinitely many states — but you can only extract one bit of classical information from it. Quantum weirdness, meet information limits. #QuantumComputing #QuantumInformation

Exploding Gradients

@ExplodingGrads

. 9. 15.

Learning quantum computing? John Watrous has you covered with a full 16-lesson course — videos + notes — on: 🔹 Quantum information theory 🔹 Algorithms (Shor, Grover) 🔹 Error correction (CSS, Toric, etc.) Free, rigorous, and crystal clear. Link: arxiv.org/abs/2507.11536

Exploding Gradients

@ExplodingGrads

. 9. 13.

Neural nets learn functions by fitting smooth, low-frequency parts first, then detailed high-frequency ones — a phenomenon called spectral bias. Early stopping leverages this to reduce overfitting. Here’s a quick visual explainer! #ML #DL #bias

ExplodingGrads's tweet image. Neural nets learn functions by fitting smooth, low-frequency parts first, then detailed high-frequency ones — a phenomenon called spectral bias. Early stopping leverages this to reduce overfitting. Here’s a quick visual explainer! #ML #DL #bias

Exploding Gradients

@ExplodingGrads

. 9. 12.

What’s really happening: Dropout ≈ variational Bayesian inference. It approximates a deep Gaussian process — giving you uncertainty for free!

Exploding Gradients

@ExplodingGrads

. 9. 11.

Quantum ML doesn’t crash. It exists in a superposition of working and not working until you try to deploy it. #QuantumComputing #MLproblems

Exploding Gradients

@ExplodingGrads

. 9. 11.

Classical ML: Takes hours to train. Quantum ML: Takes milliseconds to confuse you. #ML #AI #DL #QuantumJokes

Exploding Gradients 님이 재게시함

Rohan Paul

@rohanpaul_ai

. 9. 9.

Fei-Fei Li (@drfeifei) on limitations of LLMs. "There's no language out there in nature. You don't go out in nature and there's words written in the sky for you.. There is a 3D world that follows laws of physics." Language is purely generated signal.