- Видео 92
- Просмотров 324 202
Machine Learning Explained
Канада
Добавлен 6 авг 2018
Hi there!
I’m a PhD student at McGill specialized in computational neuroscience. I use machine learning technique to gain insight into the brain correlate of consciousness and pain. I work mainly with Python, Matlab and R for my data science project. I use React and Django or Flask for my side-projects.
In this channel, I strive for explaining machine learning concepts by implementing them using from scratch using various programming language like Python, Java or C++. The goal of this channel is for you the viewer to understand these concepts and to demystify the complexity of artificial intelligence.
Here are a sample of the explained concepts that you can find in this channel:
- Batch Gradient Descent
- Linear Regression
- Multivariate Linear Regression
- Fully Connected Feed Forward Neural Networks
- The Perceptron Algorithm
Hope you learn something here! If you have any questions you can write them in the comment section of the videos or send me an email!
Best,
Yacine Mahdid
I’m a PhD student at McGill specialized in computational neuroscience. I use machine learning technique to gain insight into the brain correlate of consciousness and pain. I work mainly with Python, Matlab and R for my data science project. I use React and Django or Flask for my side-projects.
In this channel, I strive for explaining machine learning concepts by implementing them using from scratch using various programming language like Python, Java or C++. The goal of this channel is for you the viewer to understand these concepts and to demystify the complexity of artificial intelligence.
Here are a sample of the explained concepts that you can find in this channel:
- Batch Gradient Descent
- Linear Regression
- Multivariate Linear Regression
- Fully Connected Feed Forward Neural Networks
- The Perceptron Algorithm
Hope you learn something here! If you have any questions you can write them in the comment section of the videos or send me an email!
Best,
Yacine Mahdid
What to Learn for Deep Learning? | 3 Roadmaps for Beginners
If you are a total beginner in deep learning and you are motivated, check out this video to get started with an efficient learning framework!
# Table of Content
- Introduction: 0:00
- Usual Roadmap: 0:56
- Issue with Said Roadmap: 1:43
- Best Roadmap: 3:31
- Set a Goal: 4:40
- Step 1: 5:40
- Step 2: 6:58
- Step 3: 8:22
- Machine Learning Engineer Roadmap: 10:11
- Researcher Roadmap: 12:48
- AI Product Roadmap: 15:52
- Conclusion: 18:34
Cool notebooks from the creator of Fast.AI you should check out:
📌 Image Recognition of Birds: www.kaggle.com/code/jhoward/is-it-a-bird-creating-a-model-from-your-own-data
Do check out this playlist to get started with the learning framework:
📌 ruclips.net/p/PLKy51kKbOLlz2...
# Table of Content
- Introduction: 0:00
- Usual Roadmap: 0:56
- Issue with Said Roadmap: 1:43
- Best Roadmap: 3:31
- Set a Goal: 4:40
- Step 1: 5:40
- Step 2: 6:58
- Step 3: 8:22
- Machine Learning Engineer Roadmap: 10:11
- Researcher Roadmap: 12:48
- AI Product Roadmap: 15:52
- Conclusion: 18:34
Cool notebooks from the creator of Fast.AI you should check out:
📌 Image Recognition of Birds: www.kaggle.com/code/jhoward/is-it-a-bird-creating-a-model-from-your-own-data
Do check out this playlist to get started with the learning framework:
📌 ruclips.net/p/PLKy51kKbOLlz2...
Просмотров: 354
Видео
What are Pooling Layers in Deep Neural Networks?
Просмотров 4,9 тыс.17 часов назад
You might be wondering, why we even need to do pooling in the first place for CNN. In this tutorial, we'll answer this question by exploring the three ways in which pooling is helpful for training deep vision models. # Table of Content - Introduction: 0:00 - 3 reasons to use pooling layers: 0:29 - What is pooling: 0:50 - pooling for dimensionality reduction: 3:48 - pooling for translational inv...
Network in Network Deep Neural Network Explained with Pytorch
Просмотров 433День назад
Network in Network (NiN) is a very influential paper that introduced two massively useful concepts to the next generation of deep neural network architecture. Namely, 1x1 convolution throughout the network and global average pooling for the classification layer. # Table of Content - Introduction: 0:00 - 1x1 convolution: 0:49 - Global Average Pooling: 2:33 - Datasets: 3:22 - Results: 3:43 - Code...
How to Set Up your Deep Learning Environment?
Просмотров 139День назад
Getting started with deep learning doesn't have to be complicated. The best way to set yourself up is to go on Colab or Kaggle and make your analysis from a notebook. # Table of Content - Use Online Notebooks: 0:00 - Main Benefit of Notebooks: 1:14 - Notebook Anatomy: 1:53 - Pros of Notebooks: 5:52 - Cons of Notebooks: 6:09 - Conclusion Use Notebooks: 7:29 Honestly, getting started is the one s...
Highway Networks - Deep Neural Network Explained
Просмотров 23114 дней назад
Highway Networks are a type of network inspired by LSTM that make use of learnable information highway to let inputs flow unimpeded to subsequent layers. They have a lot of similarities with residual neural networks and offer deep insight into how to make training deeper neural networks possible. # Table of Content - Introduction: 0:00 - Degradation Problem: 0:38 - Idea Behind Highway Networks:...
What are 1x1 Convolutions in Deep Learning?
Просмотров 2,4 тыс.14 дней назад
You might have come across 1x1 convolution in deep learning architecture and wondered why they were there. In this tutorial, I'll walk you through their usefulness and how they compare to other dimensionality reduction techniques. # Table of Content - Introduction: 0:00 - 1x1 in networks: 0:11 - Convolutions: 1:40 - How to reduce dimensionality: 2:48 - What is 1x1 convolution doing?: 4:25 - Poo...
![Inception Net [V1] Deep Neural Network - Explained with Pytorch](http://i.ytimg.com/vi/x9YkGOPXGcg/mqdefault.jpg)
Inception Net [V1] Deep Neural Network - Explained with Pytorch
Просмотров 34821 день назад
GoogLeNet or Inception V1 is an interesting 22-layer deep neural network that made heavy use of 1x1 convolution in order to construct a competitive deep network in 2014. In this tutorial, we’ll take a look at it’s usage in Pytorch with CIFAR-10, a step-by-step explanation of the paper “Going Deeper with Convolutions” as well as a walkthrough of the Pytorch implementation. #Tabl of Content - Int...
FractalNet Deep Neural Network Explained
Просмотров 51328 дней назад
FractalNet is an alternative to residual neural networks that was built to showcase that it wasn't the residual aspect of resnet that made it so powerful. Indeed, the FractalNet paper was able to show that it was the training of sub-paths of different lengths that made the network much more performant than VGG. In this tutorial, we'll go through the paper and a Pytorch implementation of Fractal...
What is VGG in Deep Learning?
Просмотров 2,7 тыс.Месяц назад
Depth in neural networks is a very important parameter. Before ResNet, the VGG network was able to prove its importance by scaling up AlexNet to 16-19 layers. In this tutorial, we'll take a look at the theory behind the architecture as well as a Pytorch implementation from the official documentation. # Table of Content - The Importance of Depth in Neural Networks: 0:00 - VGG Network Architectur...
Stochastic Depth for Neural Networks Explained
Просмотров 1,1 тыс.Месяц назад
Stochastic depth is a powerful regularization method applied on residual neural networks that speed up training and enhances test performance. In this tutorial, we'll go through the methodology and a Pytorch implementation. # Table of Content - Introduction: 0:00 - Background and Context: 0:30 - Questions: 2:46 - Architecture Changes: 3:00 - Why use the resnet architecture for stochastic depth?...
DenseNet Deep Neural Network Architecture Explained
Просмотров 1,7 тыс.Месяц назад
DenseNets are a variation on ResNets that swap the identity addition for concatenation operations. This has many benefits, mainly better performance for smaller parameter sizes. In this video, we'll discuss this architecture and do a full Pytorch implementation walkthrough. Table of Content - Introduction: 0:00 - Background and Context: 0:26 - Architecture: 3:25 - Data Set: 7:29 - Main Results:...
How to Run ML models on the Browser?
Просмотров 1592 месяца назад
Running ML models in the browser is now pretty easy with Transfomer.js. In this tutorial, I walk through a very simple JS app that does object recognition client-side! Code: github.com/yacineMahdid/transformer-js-test-app # Table of Content - Introduction: 0:00 - Example App: 0:32 - index.html: 1:44 - style.css: 2:42 - index.js: 3:20 - file upload event listener: 5:07 - detect function: 6:02 - ...
ResNet Deep Neural Network Architecture Explained
Просмотров 2,1 тыс.2 месяца назад
ResNets are a very useful deep neural network architecture that excel at computer vision. Here we will discuss the core insight from the architecture and look at a Pytorch implementation! Table of Content - Introduction: 0:00 - Background and Context: 1:18 - Data Set: 5:28 - Architecture: 6:12 - Main Results: 10:41 - Limitation: 12:58 - Pytorch Walkthrough: 13:58 - High-Level Pytorch API: 15:12...
Img2Vec and UMAP to Visualize High Dimensional Data
Просмотров 4152 месяца назад
A very cool visualization trick that I’ve used recently for vision classification: image → embedding →UMAP projection→ dynamic visualization 💡 Notebook for this project: 📌 www.kaggle.com/code/yacine0101/aerospace-engine-defect-detection # Table of Content - Introduction: 0:00 - Project Overview: 0:30 - Data Overview: 2:45 - Getting Embeddings with Img2Vec: 4:12 - Reducing Dimensionality with UM...
Cosine Similarity for Data Science Tutorial
Просмотров 4193 месяца назад
Cosine similarity is a measure of how two multi-dimensional data points are alike. We'll explore the metric along with a Python example in this tutorial. # Table of Content - Introduction: 0:00 - Context: 0:24 - Example of Cosine Similarity Usage - Graph: 1:27 - Example of Cosine Similarity Usage - NLP: 2:44 - Theory: 3:45 - Formula: 5:15 - Kaggle Notebook Overview: 6:34 - Python Implementation...
LLMs Learn Tools at 775M Parameters!
Просмотров 7245 месяцев назад
LLMs Learn Tools at 775M Parameters!
Solving Boggle using AI : Dynamic Programming + Trie in Python
Просмотров 64211 месяцев назад
Solving Boggle using AI : Dynamic Programming Trie in Python
What is Data Science? - Introduction to Data Science with Python Workshop
Просмотров 372Год назад
What is Data Science? - Introduction to Data Science with Python Workshop
How to Solve Programming Problem | 3 Steps Framework
Просмотров 369Год назад
How to Solve Programming Problem | 3 Steps Framework
How to Ask for Help in your Data Science Project!
Просмотров 2472 года назад
How to Ask for Help in your Data Science Project!
Shannon Entropy from Theory to Python
Просмотров 5 тыс.2 года назад
Shannon Entropy from Theory to Python
Handling Negative Result in Data Science Project!
Просмотров 3402 года назад
Handling Negative Result in Data Science Project!
Fact Check and Document All Data Science Assumptions ✅
Просмотров 1332 года назад
Fact Check and Document All Data Science Assumptions ✅
Kolmogorov Complexity With DNA Data in Python
Просмотров 8992 года назад
Kolmogorov Complexity With DNA Data in Python
Distance Metrics Explained and Visualized in Python
Просмотров 2,4 тыс.3 года назад
Distance Metrics Explained and Visualized in Python
Why you shouldn't use K-Fold Cross Validation.
Просмотров 4463 года назад
Why you shouldn't use K-Fold Cross Validation.
K-Nearest Neighbor from Scratch in Python
Просмотров 2,3 тыс.3 года назад
K-Nearest Neighbor from Scratch in Python
How to Structure Data Science Project
Просмотров 6 тыс.3 года назад
How to Structure Data Science Project
Too underrated video! Thank you so much for this video, now my vision is clear and I have set my goal - ML Engineer Job. Will soon comment here back, when I get a job/internship!
Cool stuff, do you have a target company in mind? I could make you a more detailed roadmap.
I still don't get why it's supposed to help translational invariance. As you say, the convolution is already capable of that. If I move the image contents 50 pixels to the right, the features should also move 50 pixels, given stride 1 and padding. Exactly the same is true for traditional sobel edge detectors. The edges don't change if I convolve the image with the edge detection filters translated or not
No the convolution help with translational equivariance, but not with translational invariance. The idea of adding pooling is that you are forcing exact nearby pixel/feature information to be “lost”. This means that the network is forced to learn more generalizeable components of the picture you are showing it.
Hey, great video!! Do you go to university in montreal? I'm from ottawa and will be going to Waterloo for computer engineering.
Hey there, I was going at McGill yes! Cool stuff, best of luck in your computer engineering journey :)
You made this concept so easy to understand: thank you!
I am glad it was useful :)
Very useful indeed, I just think the 3D representation could be a bit better (I'm used to seeing the filter rectangle behind the first layer and not on the side, but that's probably just me)
True, it’s titled by about 90 degrees. Otherwise the 1x1 convolution wouldn’t fit well in the image I believe.
Is this code for works only for same size of images ? Because it was giving while i used for my images
Hey there, it’s been a while I’ve checked this code. What were your image size?
Thank you for breaking it down so well! Keep up the excellent work!
Hey thanks for the kind words! Will do, I have a few more videos ready for next week :) Do let me know if there is a specific topic or question you would like covered!
I like your explanations, but Im watching it randomly As I'm beginner, where should I start?
Glad you liked it. It’s the second time this week I had this request for where to start in deep learning, I’m setting up a video on that topic will publish soon!
Shouldn't we call it a 1 x 1 x Cin convolution?
That would indeed be a less confusing name for sure. That thing already have like 7 different names though haha
Thanks a lot for the content. One small request. Can you pls reduce the background music volume next time?
Hey there, Yes sorry for the inconvenience, I've improved the sound in my new videos :)!
Very interesting! Thank you so much! :)
My pleasure, glad it was useful :)
I am totally new to machine learning. From where do I start?
Great question, will make a video on this next week!
Thanks , how can we use VGG for 1D signal? Is it possible to use VGG for regression instead of classification, how?
Hmmm depends, what’s the 1D signal about? Is it visual?
It’s acoustic signal.
Ah then no, VGG shouldn’t be your pick here. It was expressively designed for image classification. Take a look at the various model on PyTorch made specifically for audio signal: 📌 pytorch.org/audio/stable/models.html
Can’t we change output layer, activation function to do regression?
Yes you can, but the internal of the model is tailor built for image. If you are able to express your 1D signal input as an image I would say it might be worth it to try. However, there are other models made specifically for audio.
very cool thank you for the knowledge
Glad it was useful! :)
Glorious😳
😀
Hey, fyi I had to reshoot some of the section on this video because I couldn’t stop saying Drop Path (from Fractal Net) instead of Stochastic Depth. There is still 1 wrong mention of drop path in there that I wasn’t able to fix haha That's what you get from reading two paper simultaneously!😅
thank you very much for your nice work can i have your slides?
Thanks, for sure! You can shoot me an email at mail@yacinemahdid.com and I'll send them to you.
Thank you! Exactly what I needed.
Glad it was useful :)
Nice work, but why was the first file overall a larger file to start? Character quantity?
Yes, but this shouldn't have happened. The first file contains 2M characters, while the second only 1M thus |file_2| = 2 * |file_1|. The author of the video wanted to have a first file with 500'000 "ab" (thus 1M chars) but that's not what he did.
Amazing video! Thanks!
Glad it was helpful! :)
Who the heck is Tanimoto
This is SUPER helpful. I was looking online for a good example using sklearn to no avail. Even asked ChatGPT and was led astray.
Thank you! this was exactly the explanation I needed
Glad it was helpful, any other topic you would like me to cover?
Will please tell me how to do ndam optimization in matlab for deep learning
Hey there! For sure, do you already have some code that I can take a look at ? Also, why are you using MATLAB?
thank you so much it helped me in project lot
Glad it was useful :)
can you give me the link to the exact jupyter notebook? I can't find the exact same jupyter notebook from the link that you provide on the description
Appreciate it man thank you. This is golden.
Glad it was useful, keep it up !
Thank you for your concise and clear explanation!
Glad it was helpful! This tutorial was pretty niche haha
If I'm not mistaken you only return the entropy for the last checked nucleotide: entropy = 0 for nucleotide in {'A', 'T', 'G', 'C', 'N'}: rel_freq = dna_sequence.count(nucleotide) / len(dna_sequence) if rel_freq > 0: entropy = entropy - (rel_freq * math.log(rel_freq, 2)) return entropy
hi
Hello 👋 Let me know if you have questions!
@@machinelearningexplained Do you know a way to use the MNIST dataset without using sklearn, pytorch, or tensorflow? If not, what are some datasets that you recommend?
This is a great project for beginners, thanks for this valuable information !
Glad it was interesting! I remember the first time I've attempted to make a boggle solver, my DFS based algorithm was extremely inefficient. I had to build it in C and have parallelization in place to barely hit the 2min mark for the game. Once you use dynamic programming and a search-optimized trie structure it's a world of difference.
Hi, can you share example on fine and gray modeling? Also, where I can get these codes? thanks!
Hey there, Sorry for the wait this comment slipped through! What do you mean fine and gray modeling?
promo sm
What?
Love this video !
Hey Yacine, glad you liked it!
Where are you from?
I'm originally from Algeria, but I've lived in Canada for most of my life! btw you can join the discord for general casual chat : discord.gg/QpkxRbQBpf Easier for me to follow up!
Keep going 💪💖
Thank you, I will
Welcome back
😁 it's been a while haha
You're back again! That's amazing ✨
Yes I am! I have a few videos lined up :)
Can you please help me with rainfall problem with deepxde and physics informed neural network
Yes sir for sure, shoot me a message on LinkedIn I have some time tomorrow for a video-call 📌 www.linkedin.com/in/yacinemahdid/
@@machinelearningexplained please accept my request, and thank you sooooooooo much for replying me
So for nesterov gradient, the learning rate would have to be set constant and not found automatically using line search for each iteration?
Great video. Thank you
Glad you liked it!
Great video. Thank you
Thanks :)
Very helpful. Thanks!
Glad it was helpful Kevin!
Nice tutorial, thanks a lot. But! If you explain something - just explain. Watching how you was moving the pieces of code was so annoying!
Hi there, Thanks for the feedback! Will improve that part in the next tutorial for sure :)
i'm so frustrated that i don't know *hope i can learn to code math formulas in future* if anyone got some guidance please give
Hey there Jonas, What do you mean? Is there a math formula in particular you are struggling with?
thank you very much for the tutorials and code! But I don't quite understand why both AdaGrad and AdaDelta perform poorly for these examples?
Hey there, Glad it was useful! The example use a very very basic formula with which AdaGrad/Delta are way too overkill for. It was more to illustrate that we could code these formula and they can work in practice. To make them work well I would have had to tweak the hyperparameters some more. In a neural network though they are good optimizer!
@@machinelearningexplained thank you very much, I used your work to create a colab file in which I animated the different algorithms exposed :)
@@Lorenzo8690 wow cool, shoot me a link I'll check it out!