- 21
- 122 035
Daniel Krei
Lithuania
เข้าร่วมเมื่อ 26 พ.ย. 2018
The Idea Behind Optimizers
This is my take to give you a high level explanations of how optimizers work as fast as possible.
In this video we are looking at gradient descent as an example. In the folowing videos we are going to take a more in-depth look into how different optimizers work.
In this video we are looking at gradient descent as an example. In the folowing videos we are going to take a more in-depth look into how different optimizers work.
มุมมอง: 9
วีดีโอ
NEW AI video tool coming soon
มุมมอง 4.9K2 หลายเดือนก่อน
I've been experimenting with a lot of generative AI tools and thought about putting it into a small saas product for a while. If you have any thoughs about what is lacking in the market - let me know! Text-To-Image and Image-To-Video pipelines would be available. At the moment, I want to know how many of you would like to see it live. Here is a link to a landing page: atarahlabs.com Be sure to ...
The Idea Behind Neural Networks
มุมมอง 4.5K2 หลายเดือนก่อน
This is a quick video explaining the main idea behind neural networks. Why we need them, how and why they work. Although given samples are simple, I believe they will let you gain the intuition about what happens inside neural networks as they learn.
Matrix Multiplication Explained | Math for Machine Learning Part 5
มุมมอง 2723 หลายเดือนก่อน
This is a short introduction into matrix multiplication. Matrix multiplication is the fundamental linear algebra operation. This is the 5th video in a series about mathematics used in machine learning. Dot product: th-cam.com/video/5-xLHOMekNk/w-d-xo.html
Transpose of a Matrix Explained | Math for Machine Learning Part 4
มุมมอง 1594 หลายเดือนก่อน
This is a short video about the transpose of a matrix. Part 4 in series about mathematics for machine learning. Python and Numpy examples are given with visualizations of transposing matrices and vectors.
Dot Product Explained | Math for Machine Learning Part 3
มุมมอง 1294 หลายเดือนก่อน
This is the 3rd video in a series about mathematics for machine learning. This time we are looking at the dot product. A very simple but fundamental linear algebra operation. Part1 (scalars, vectors, matrices and tensors): th-cam.com/video/WpxKkvX9kgc/w-d-xo.html Part2:(simple operations): th-cam.com/video/yHAMuOAcm6c/w-d-xo.html
Simple Operations with Vectors and Matrices | Math for Machine Learning Part 2
มุมมอง 3554 หลายเดือนก่อน
This is the second video in this series about mathematics for machine learning. This time we check basic operations we can do with scalars, vectors and matrices.
Intro to Scalars, Vectors, Matrices and Tensors | Math for Machine Learning Part 1
มุมมอง 7734 หลายเดือนก่อน
This is a short introduction into scalars, vectors, matrices and tensors. This is the first video in a series about mathematics used in machine learning. In the next video I’ll talk about basic operations we can do with them and I’ll share some code samples.
Oppenheimer AI trailer (Runway Gen2)
มุมมอง 8889 หลายเดือนก่อน
This a little generative AI experiment using Runway Gen2 to recreate Oppenheimer trailer using only text prompts.
Precision and Recall | 3 Minute Tutorial
มุมมอง 2329 หลายเดือนก่อน
A quick intro to Precision and Recall. These are extremely important metrics when analysing machine learning models. If you want me to explain other metrics, leave a comment!
IoU (Intersection over Union) | 2 Minute Tutorial
มุมมอง 40610 หลายเดือนก่อน
This is my attempt to explain IoU as fast as possible. Mathematical animations are made using Manim Python library. Images used in the video are from this dataset: www.kaggle.com/datasets/hngngn/portrait-segmentation-128x128
Understand Cosine Similarity | 2 Minute Tutorial
มุมมอง 7K10 หลายเดือนก่อน
This is a quick introduction to cosine similarity - one of the most important similarity measures in machine learning! Cosine similarity meaning, formula and example! If you like this video, hit a like and subscribe! Paper icon used in a video: www.flaticon.com/free-icons/paper" -Paper icons created by monkik - Flaticon
Learn Euclidean Distance | 2 Minute Tutorial
มุมมอง 1.5K10 หลายเดือนก่อน
Euclidean distance is probably the most well-known distance measure. Here is how it works! Animations are made using Python Manim Library.
Make a 3D Neural Network in Blender! | Code
มุมมอง 2Kปีที่แล้ว
This is a 3D Neural Network visualisation Blender code walkthrough. It shows how I made logistic regression part of neural networks visualisation video (video link: th-cam.com/video/hIYR6qMXujE/w-d-xo.html) All training code, required files and Blender script available here: github.com/DanieliusKr/neural-network-blender
4D: Easy Peasy 5D: My Brain Is Burning
Brilliantly concise explanation. Have a like. :)
My trying to untangle the cables behind my computer
It may seem like “spinning”, but the entire time, your brain gets so confused so it thinks its switching. In reality, it’s just a cube in a cube. EDIT: Find the right frame? If you did, you’ll see that the cube is inside a cube.
THANK YOUUU
Xyztuvw
Like eventually, it wasn’t even a cube anymore. It just became a mangled piece of marble and mesh.
2D: square 3D: 6 squares 4D: 2 copies of 6 squares 5D: 3 copies of 6 squares 6D: 4 copies of 6 squares 7D: 5 copies of 6 squares
3D: Cube 4D: Hypercube 5D: Supercube 6D: Ultracube 7D: Megacube
I know 1d, 2d, 3d, 4d and 5d. 6d hurts my brain
Me watching 7D:💥💫💥💥
the only thing my brain can comprehend was the 3D cube, the 5D cube, and the 6D cube
Bro had a glow up
7D Cube, The Spider Web Cube.
The color of these change lime the phases of the endbos
Cubu in a cubu in cube in a cube in cube???
The 4rth dimension is why gravity exists, the fabric of space time
Me when I watch too much sci-fi at 3 am:
Too high IQ for me
It’s just more cubes upon more cubes inside of more cubes
3 dimension to 10D dimension Transitioning from three dimensions to ten dimensions involves a significant shift in our understanding of space. In higher dimensions, the geometry and properties of space become more complex and difficult to conceptualize. In theories like string theory or M-theory, which propose extra dimensions beyond our familiar three, these additional dimensions are often described as compactified or curled up at extremely small scales, making them imperceptible on everyday scales. These theories suggest that these extra dimensions could play a role in explaining phenomena such as the behavior of fundamental particles or the nature of gravity, but their exact implications are still a subject of ongoing research and debate in theoretical physics.
Duuuuude that is highly appreciated kinda word m telling you I have been trying to figure these out in the initiative way from the last 3 days and I was able to explain it only. Not I can make it understand to the person who is asking woooowwww😮
I dont want to try and figure out what was going on.
Is the code availabll for this?
Now show a 0 dimensional cube
This is a mega cube made with 10 hypercube
POV:your brain while working
I'm so confused to the point that my face is stuck like "😀". I think my braincells died 😀 help 😀😀
All of the are my blanket when I try to find the long side
4D cube in a 3D engine on a 2D screen made out of 1D lines
Indeed pretty amazing explaination, helped me a lot! Thanks.
Among the variations, the hexagon and square always appear in all dimensions
cute
obviously the 4d one would be a hypercube so i could name the 5d one, a hetratesseract
all of them are n-hypercubes: 0d is a point, 1d is a line, 2d is a square, 3d is a cube, 4d is a tesseract (because tessera = four) and from then on you have 5d penteract, 6d hexeract, 7d hepteract and so on
Is this project/ addon available?
not at the moment, but I have plans to make a tutorial
I think 3 is the optimal number of Ds
A 4D, 5D, 6D, and 7D cube, in a 3D plain, projected on a 2D surface, for my 1D brain to handle.
And my 0d a** had a stroke reading whatchu said (I understood it was just to fill in the reply)
The peaceful music with the demonic cubes on screen fits perfectly!
evolving to the space age -
I just saw a pyramid, 7 cubes, physics folding in on a fold out, a cube rotating in both states until observed, Heisenbergs uncertainty principle, a tesseract, PI, A pentagon, Hexagon, Octagon, Triangle, Cube, Square, Cone, hypersphere, Gravity, The eletromagentic, Weak and strong nuclear force, The fabric of space time, the cosmic microwave background radiation, quantum flucations, Zipfs law, a Singularity, And my brain got destroyed
Totally cubular man! 😎
Cool stuff
.....
Looking good. However the 5D hypercube is missing a few edges. At 0:39 you can see that big gap and that the vertices surrounding that gap only have 4 edges connected to them. At 0:47 you can see it even better. Sort of like you only rendered 9 of the 10 hypercells. The 6D and 7D cubes seem to be missing even more edges as you can see at 1:01 and 1:23. In general an n dimensional cube should have n edges connected to each vertex with no gaps. Either way i really like that you actually make use of all the rotational planes of the n dimensional space. Out of curiosity: Do you project the dimensions down one by one? Like 7D -> 6D -> 5D and so on or so you project down from 7D straight to 2D while using the pythagorean theorem to calculate the distance? Because of the 7D cube some vertices go out quite far at some points.
Thank you for your comment! Yes, you're right there are some missing edges. And yes, I project them one by one down to 3D. It is done using blender python api.
Wow, that's a cool project! By the way, is the input.npy and output.npy in github only represent one situation? What should I do to have a full animation from 0 to 9? Thx a lot!
❤
That was good. Thank you.
the hexeract is too hard
👏🏻👏🏻 awesome explanation
Why do we need matrix multiplication?
"It hurts my brain just to think "it's just spinning"" -Someone