Make Your First AI in 15 Minutes with Python
Вставка
- Опубліковано 28 січ 2021
- Make your first AI using Tensorflow/Keras and scikit-learn. This AI model is trained on real data from breast cancer diagnosis. Within a couple of minutes, the AI can classify tumors on its own!
Get the code:
colab.research.google.com/dri...
Get the dataset:
gist.github.com/adameubanks/3...
If you want to learn more about neural networks and machine learning, consider watching my two tutorials:
Machine learning with scikit-learn: • Learn Machine Learning...
Neural networks with Keras: • Learn Keras: Build 4 D...
Get 60 Free Days of Unlimited Access to Scribd!
www.scribd.com/g/6lchwh
Follow Me:
Twitter: / khanradcoder
Instagram: / thekhanradcoder
GitHub: github.com/KhanradCoder - Наука та технологія
This was an amazing example of clear and concise education. You are nailing these tutorials better than some 60 year old professors. And with push-ups.🤯
Literally the first module given by coursera, sure it's nice to have something to put on 2x speed but it's not what I was looking to learn
Quantum computing could threaten humanity as we know it…
ua-cam.com/video/ip3FwAYWjkw/v-deo.html
In this Video you learn NOTHING about AI and you dont even do anything in general why im giving it a thumb down.
@@ElmoSeesYou so you're saying that Tensorflow he's using with neural network prediction isn't AI at all?
@@nan6849 I don't know what happens in the module, but I believe it actually is AI. But what I'm saying is, that you just don't get to know anything abiut AI. What even is a neural network? No clue because this video is explaining nothing
If you get an error at the end that is something like "ValueError: Input 0 of layer "sequential_1" is incompatible with the layer: expected shape=(None, 455, 30), found shape=(None, 30)" then replace the 1st line the segment "model.add" instead of "model.add(tf.keras.layers.Dense(256, input_shape=x_train.shape, activation='sigmoid'))" use "model.add(tf.keras.layers.Dense(256, input_shape=x_train.shape[1:], activation='sigmoid'))". Most importantly recompile all segments both from start to finish after making this change.
omg recompiling i did not do that thank you!
Thanks man, it worked!!
@@thehulk6866 your welcome 😀
Thanks for sharing! Can you explain why the error was happening in the first place? Is it because the x_train shape is still including the first column and hence this solution is just slicing the data set from to start from 1 instead of 0?
Hey gem! THANKS FOR SHARING 💎❤️
Perfect video! BTW for those experiencing an error at the end with the input_shape giving ValueError: Input 0 of layer "sequential" is incompatible with the layer: expected shape=(None, 455, 30), found shape=(None, 30),
Just use this instead of the first line:
model.add(tf.keras.layers.Dense(256, input_shape=x_train.shape[1:], activation = 'sigmoid'))
Remember to run all code blocks from top to bottom again, to make sure it works perfectly fine! Do not try to just run this separate code block otherwise it will never work.
Thank you! I was stuck at the same spot.
What are differences between them?
@@user-mw1uk9sn9w no difference, its just a colab bug nowadays that most machines cant handle for some reason. the second line basically removes all bugs and when u run it all in a row it makes sure that the last line is executed properly
TYSM this is the comment I was searching for
You rock. I've been scouring the internet for this and it was right there x)
Thanks! I liked the part where you opened the internet tab to show the parts of the neutrale net in reference to the code. When I see code I usually get flustered but that illustration helped make sense of the code. Please keep making more videos! 😀
If you got it to work could you tell me how? I couldnt get past the "model.fit(x_train, y_train, epochs=100)" Line. If you did it, please let me know how!
@@nik7069 read @Saad Humayun comment, it saved me
@@didyouknow19256 ab iska comment kaha khojne jaun 🙄
@@didyouknow19256 where is it
It's not visible
Could u pls tell me about it
Great video. As someone who as taken a full AI course in College I can say that they should really use your example as lecture 1. Show the students what they can do, and then teach them each of the components that they simply used from the tf library.
As a college professor for years, a good approach (depending on the starting points of the class) would be to give the students this video and a few other bits of material before class, get them building themselves, and then help with errors and confusion during class time.
@@InconvenientxWhat would you say those other bits of material are?
College Protest Preparing
Thank you for this awesome video Kahnrad. This is wonderfully explained.
UPDATE: model.fit() will throw an error unless line 24:
model.add(tf…Dense(256, input_shape=x_train.shape, …)
is changed to: x_train.shape[1:]
Then recompile top down, should be solid
Ty! I was stuck on exactly this
Thank you!!
Thanks!
It would have been a frustrating way to spend time if I didn't see this...
the accuracy is 0.0000e+00 tho. Idk how to fix that
Thanks for such a comprehensive and informative guide! I’m trying to make an AI that manages data from plants in my garden, and this has been such a great help.
Breath of fresh air. Crystal clear and to the point. Thank you so much!
Great, great video. Using the Tensorflow & Keras with the scikit-learn was super cool, especially in knowing it will help doctors make important cancer diagnoses, makes it awesome. Great job!!
Very clear now I been looking for stuff about this for three hours and you made it nice and clear amazing video :)
Great video, very clear. Any ideas about the warning at the end of the program about incompatable shapes between the model and the input? Obviously it works, but is this something that needs fixed in a more advanced application?
GOOD Job Man! Love how humble you are!! Good Job!!!!!
Not sure how or why I came across this video, but muhfkr you know what you're doing on a comp!!! Good job!!!
Good stuff young man. Very easy to follow and straight to the point. Appreciate you
Fantastic videos! Very easy to follow and actionable. Will absolutely share with everybody interested in this field. One quick note, when, doing anything in the medical field you have to understand the truth labels of the data.
Yo i think OpenAI copied your tutorial
no
fr?
@@theadameubanksno dawg its a joke
@@CoolDude-mq8dh he knows its a joke he just played along with it
This video may become part of their training dataset, so eventually maybe yes.
Thank you just learned programming or understood more and more the basics in c# and was looking to use ai in my job in a production company. Your video was a huge help considering I did python last I think 7 or 8 months or even longer but loved the simplicity of that kind of subject. I know it can go complexer but the basics are super easy to learn because of you! So thank you and keep up your work man! :)
Thank you!
You are AWESOME! You are a prodigy at coding. Please continue what you do!
thank you man, this seems so easy to understand. thanks again. great job.
absolute legend, im in uni and you explained it better than my lecturers aha
This was great! I am a clinical psychology and am I'm helping with retention in the treatment program for at risk youth. Babe I've been looking for a way to do to do machine learning to predict which kids are at highest risk in order to throw more resources behind them. I think I've found it! Thank you! If this works the way that I hope it will then you will be impart responsible for adding to do the safety and securityOf hundreds of youth.
good luck !
Well? How'd it go? You've got to tell us 😅😂☺️
I would love to hear an update on this.
pretty crazy nail the kids just going into high school with AI predicted system.
I'm curious about this and maybe diagnosing cars and certain model issues. Or even install a system that helps tell what's causing issues.
My dude, this was very helpful for my research! Thank you!
Very clear , easy and straight to the point. very good explanation
Bro, I just finished watching this 2022 July... And I really appreciate your efforts so much.
For people getting the ERROR at the end use this line instead
model.add(tf.keras.layers.Dense(256, input_shape=x_train.shape[1:], activation='sigmoid'))
on the first line when adding models
i m still getting that error. can you help??
@@pratikbehera4770 did u know how to fix it? I’m also still getting it
@@fufusytb nah...tried some of the more comments still getting the same error
@@fufusytb i put this code on the last line and it worked, but the data doesn’t configure with the gpu for the testing of the AI machine we just created
For me it worked with:
model.add(tf.keras.layers.Dense(256, input_shape=(x_train.shape[1],), activation='sigmoid'))
thx man i have just started my project which is related to deep learning and this helped a lot
This video is very good and you made the whole concept very easy to grasp beside that you have a strong good voice that keep me concentrating with you through the whole video, thank you.
Glad you liked it!
good job. the best teaching is when you make it simple.
Thank you for your contributions to society and the internet with your free information, Brother.
Very interesting, now I know something I don't know before. Keep posting and thanks.
This dude needs more subs he has the best content ever.
Great tutorial and the example data set is a perfect example of why AI is so cool!
This AI is great, and the video is perfect. For those experiencing errors in the last two blocks, here is what to do: Change the fifth block's code to this:
import tensorflow as tf
model = tf.keras.models.Sequential(tf.keras.layers.Dense(1))
Run the fifth block a few times, then try again- it should work
Thanks for this, this is the only code that worked with mine so you're a lifesaver
Thank you, that is very helpful indeed!
You are awesome!! it worked!!
thanks a lot man this helps a lot! may i know why do we need to code as you mentioned and how does this fix the error? im still new to machine learning and trying to get to the bottom of this.
Thanks, this made it work. Probably something has changed in the syntax in last 2 years.
Can you please upload more for AI tutorials? Great video btw
Yes same here
Same
Yes , more A.I. tutorials please 🥺
This is such a amazing tutorial im dumb with learning Coding i just started but i was able to understand this completely im definitely sticking around
Thanks for your comment! What types of tutorials would you like me to do next?
loved it and your explanation was top notch!!
Hi @Khanrad ; can you make a video on how to mini-graph those outputs inside the IDE? So we can repeatedly visualize our output? Scrolling through the data is OK when the set it small. Ideally Colab-Jupyter would have a button to the right of the output to mini-graph upon completion. Further to that I'm aware output should be stored for CI/CD analysis later. It's a convivence request.
I use Low Code in Azure-ML, I can highlight output and it will show a mini-graph. or I can export it to PowerBI. Output in my environment is saved in the dataverse and ETL with Synapse.
I use the graph thingies with this code:
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
import tensorflow as tf
def visualize_output(output):
if isinstance(output, pd.DataFrame):
for column in output.columns:
_plot_mini_graph(output[column], column)
elif isinstance(output, pd.Series):
_plot_mini_graph(output, output.name)
else:
print("Invalid input type. Please provide a DataFrame or Series.")
def _plot_mini_graph(data, column_name):
plt.figure(figsize=(8, 4))
if data.nunique()
Thank you so much, this was very educational.
thank you so much for this:))
Fantastic class, Bro!
Thank you Very much!
You got it to work?
Broh you're legit. Keep going we are expecting more tech content with you.
very good job man I loved the video a lot. can you or any one tell me what is the operating system you used in the video?
this dude made this tutorial and did push ups lmao but still THANK YOU FOR TEACHING
Jesus Christ is God and is the only way. Hell is real whether you believe it or not. I used watch wicked anime, mastberate/porn, vape, beer, violent video games and now I don’t do that anymore; I didn’t even think changing how I am now. Hell is real whether you believe it or not. There isn’t multiple ways; catholic, muslim, etc will lead you to hell and I was a catholic before!
@@SquidBeatsYou're a bot. Whoever made you is an extremely sad human being. Attempting fear tactics to mold others to your religion goes against your religion. I'm sorry, but nobody cares. Liking your own comment is pointless. Quit attempting to force people to conform to your ideas just because you think enjoyable aspects of life are bad.
@@SquidBeats so what religion should i be to avoid hell ? protestant ? jedi ?
@@SquidBeats Some grandma died.and back from the dead.She described that hell/heaven is not real.Just a void and nothing.
I'm in ROS(robotic) religion btw.
@@deadyanothaikiropool1chait713 That's the Bible. Thank you for describing the death experience
Really like how you teach. Thanks
Hey! great video! I do keep getting an error even on your colab notebook in the line
"model.fit(x_train, y_train, epochs=1000)"
I get:
"ValueError Traceback (most recent call last)"
" ValueError: Input 0 of layer "sequential" is incompatible with the layer: expected shape=(None, 455, 30), found shape=(None, 30)"
Any thoughts on what might be happening?
man , this is really fun thank you for the good video !
Nice video bro! the push-ups part🤣
special props ;for the puships! great dude!
ur a legend u deserve all the love
Well the simp in your name isn't wrong lol
@@echotube3111 i simp for this guy
This was really helpful me! I understand they are supervised dataset and how do you train when you get some books or pdf?like notebookLLM?
Excellent video! Thank you.
Hey, I'm learning machine learning for fun and was wondering how I would be able to input new x data for the AI to generate y data for
I mean, I really don't understand much about advanced coding and algorithms like this, but it was interesting following along. Just wish I could understand more of whats going on...
Follow a yt channel called Siraj Raval. Search for his videos about activation function, loss function, etc. It taught me a lot about
AI / Machine Learning
Well done !! Thanks so much. Would you be able to advise on how to perform the AI Generation on our own computer? Running code on Google's computers is not cool AT ALL. Make you dependent, strictly no privacy, resources are limited, requires to be online . . . A great tool, but super limited
thanks for the video... btw how can i make my own data sets? i mean how can i extract tumor features from images?
Very useful video. Thanks for uploading!
how do you feed the AI data? Do you just evaluate a random x_test matrix with it and it auto classify y values?
Surely going to try this out 🙂
your teaching style is awesome
My friends, search for your life purpose, why are we here?? I advise you to watch this series 👇 as a beginning to know the purpose of your existence in this life
ua-cam.com/play/PLPqH38Ki1fy3EB-8xmShVqpbQw99Do2B-.html
Hey dude. I love your work. Can I use your example for an presentation in my university? I will give you full credit for it in my presentation.
what we need to modify if we want to the AI analyse a different csv file ?
Great video!!! thanks a lot!
Thanks for the video!
Ty, you helped me to understand Neuron
Here is the fix for the problem you are all having with the 'model.add' line . . add this to the block above it
import tensorflow as tf
model = tf.keras.models.Sequential(tf.keras.layers.Dense(1))
thank you sm
thank you buddy
Thank you
You are awesome, Finally, it's working .. Yay ! Thank You so much !
he really did the push ups LOL
Nice video, very helpfull :)
Very informative , whole lotta thanks for the post tbh
Wow I don't know much about Machine Learning I'm only thinking about learning it for now and I understood most of what you just taught! Awesome video make more like these!
Hey, just a tip for you as you go about learning. I see videos like these as a cool introduction to the subject that's more just like "ooh cool, look what that does". The mistake I often see people make is they just watch lots of videos like these, thinking they're learning. You're not really learning anything, you just know the pure basics of how a NN works. If you really want to get into machine learning, it's a lot of math and theory - so be prepared for that. If you have any questions feel free to ask :p
@@colinbalfour1834 i am currently learning ml, i do come up with some doubt whenever i work with some datasets , may i know your discord or any other social media you have
@@muhammedjaabir2609 lo yt keeps deleting my comments, whats your discord
@@colinbalfour1834 just wanting a subsubject name, if you could go back in time right now and relearn some of these data science concepts, how would you start??
and for a more specific one, i learn by creating my own mental model of 'axioms' or rules in structure as opposed to studying rigorously, so in my benefit do you happen to know of any group of topics or problems that show the majority of general rules id see in the subject? maybe moreso for beginners
@@colinbalfour1834 hey, what do you think is the best way to learn the mathematics behind neural networks? I’ve tried to learn some math regarding linear algebra and stuff but i’m a bit lost on how to apply it directly to machine learning.
I am getting the below error
Epoch 1/1000
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
in
----> 1 model.fit(x_train, y_train, epochs=1000)
Wow, this is amazing. This is my first ever attempt at building something. Quick question though, if I increase my neurons to 512, and my epochs to 2000, why would my accuracy go down?
Hi, which resolution do you recommend to watch your video?
On my tablet with regular HD-resolution I can't read the text given.
Many regards, Alex
You are very smart how you explain is clear thank you
if someone have error :
ValueError: Input 0 of layer "sequential_1" is incompatible with the layer: expected shape=(None, 455, 30), found shape=(None, 30)
Then clear whole block 6 and paste :
model.add(tf.keras.Input(shape=(x_train.shape[1],)))
model.add(tf.keras.layers.Dense(128, activation='sigmoid'))
model.add(tf.keras.layers.Dense(256, activation='sigmoid'))
model.add(tf.keras.layers.Dense(1, activation='sigmoid'))
thanks
Awesome tutorial. Next time do a fitness tutorial in between lol. Subbed
great video! loved it
Great Content! Thank you.
Wow. Thank you for this video. I’m going to put this together. How can you make the AI work with other radiographs like musculoskeletal ultrasound, MRI and the like?
Also would there be any benefit in running the AI on a quantum computer other than speed?
It was amazing thank you .. but can creat a model using more complex data set, i mean use some feature enginering and add more layers and use earky stopping method ...etc
Thank you for your knowledge Im still learning
This is helping me so much in my science fair of school😊
Thanks, it w really helpful. Could maybe zoom in a little bit so people with small screens could see the codes better, otherwise great
In your first line of model.add (where you add the X values), why do you actually take 256 ? Shouldn't this be just the number/amount of x colums in your dataset ?
The avenues for making those recognized patterns useful are diminished by illiteracy concerning the source signals and the lack of care taken in organizing the structure of the machine.
Similar to drawing constellations whereby connections are made between neighboring stars, yet those same celestial bodies are projected onto a two dimensional retina.
One might call what the Greeks did image recognition.
you got teaching skills!!
A couple of questions… why not clean the data first? (Fixing null values, duplicates, etc.) unless it was already cleaned… but also 1000 epochs might overfit the model no? The way I was taught in college was to first clean the data and always, always test for overfit. Just curious if there’s reasoning behind skipping these steps
Great job!👍🏼
that's pretty useful. You are really awesome, thank u bro.
it maybe a stupid question but how do I then input values for the AI to look at?
What site did you use for your dataset?
question: how to run this then? like will it say 0 or 1 as diagnosis?
What computer do you have? GREAT VIDEO BTW
Excellent !!! thank you
I got this: ValueError: Input 0 of layer "sequential" is incompatible with the layer: expected shape=(None, 398, 30), found shape=(None, 30)
why? how to fix it?
Can you do more AI or ML videos that could apply to daily life? Thanks man
ua-cam.com/video/CT6Zif1cLTM/v-deo.html
Hello thank you for those amazing easy videos… but i tried to copy ur method on my project but it gives errors at (model = tf..keras.models.sequential() ) what should i do
Well I’m going to follow this tutorial
hello, i was wondering how i might find the standard deviation of the accuracy?
you make big things in small... God bless.....
Great content!!
Brother great video!