Good video again !! I tell you that through your web scraping video with selenium I was able to program a bot to use it in my work test and achieve the job position thanks to that. Thanks again for teaching without any interest and without knowing that you can help people achieve their goals.
oh wow!!! I'm so happy to hear that!!! 😁😁😁 Great job on the the bot and congrats on the new position!!! I'm super happy I could help and thank you so much for your beautiful comment, Martin! 😊
Thank you so much Debayan, super happy you liked it! 😀 I had the funniest photoshoot in the world when filming for the thumbnail - I have no megaphone nor a sporty headband... so if you look at all the original photos, you'll see an Axl Rose impersonator holding a makeup remover and shouting on imaginary creatures 🤣🤣🤣 the power of graphic design is only limited by ones imagination! hahaha
Hey maria, I can't believe how incredibly awesome it was. I have never seen in my life someone train a ML model only using numpy and pandas. It's incredible, hats off!! Love from India! ♥️♥️
Masha I am quite new in this field and you are a super clear teacher. Following you it is a pleasure. A big Thank. I just wanted to get you a beer for your kindness. Ciao (I am italian!)
For those of you who got confused about why the "np.dot(feature, weights)" worked on two matrix of the dimentions 3X1 since it supposed to work as matrix multiplication. I found out that, the "np.dot(feature, weights)" will convert the second one dimentional array which is in this case the array "weights" to a column matrix. In other words, the dimention of the matrix "weights" will be converted from 1X3 to 3X1
@@PythonSimplified HELL YEAH!!! We survived August 29 1997 and we shall not let 2029 happen EVER!!! Now, when you say "lose pounds", can that mean "junk data"??? 🤣🤣🤣 🥁Siempre un gusto ver tu lecciones bella Mariya ❤ Gracias
Thanks! I tried doing this on paper during my PhD but it is just so laborious that I got bored lol! But now with Python it is so easy to run through the basic operation of the perceptron. Awesome!
I love the thumbnail hahaha. Amazing video my friend, I wasn't able to come to the premiere but I hope next time I'll be there, have a wonderful day :).
Really really thank you for this video!! First of all, your natural pedagogic system is amazing! Your english is clear, so the people like me who live in Argentina can understand all and practice the lenguage. About the step by step code, is great, i can see the importance of having a CLEAN CODE! With order, details functions, so you can see it like a dron camera! jaja This world of AI is amazing... now, please, give us some time to practice and to naturalice the concepts of these lasts videos :) would be the next one an app with AI? mmm , thanks again!
Thank you so much for the incredible feedback!!! 😊 I'm actually about to get back to this AI series from a more practical perspective 😉 We will begin diving into Pytorch and ultimately YES - even combining AI in your personal apps 😁😁😁
Brief correction: in cross entropy we use natural logs, instead of log10. Because of the activation of the network, this normalises the prediction. E.g. sigmoid and softmax use E^x, and to normalise an exp, because exp is the opposite of natural logs, we use log, not log10 since it’s not a 10^x problem.
Brilliantly done, Mariya. Do you have a theoretical reference for why the average_loss always converges to around 0.27 regardless of the number of weights or the number of training datasets? Random x's and random 0/1's is a stastically stationary system, i think, so the average_loss should always be about the same. Do you have a reference for calculating the average loss with a formula rather than by simulation?
Hi Lopo! 😀 Before you watch this one, make sure you've seen the previous episodes where I explain all the neural stuff in detail 😉: Episode 1 - Perceptron: ua-cam.com/video/-KLnurhX-Pg/v-deo.html Episode 2 - Cross Entropy Loss: ua-cam.com/video/EJRFP3WmS6Q/v-deo.html Episode 3 - Gradient Descent: ua-cam.com/video/jwStsp8JUPU/v-deo.html This tutorial connects all the dots together and in the end you'll have your very own AI project to experiment with! 😁
@@PythonSimplified yes since then , from m'y perspective i Saw that ur content was awsome when u started your Channel step by step with that python and Scraping intros, but from months ago u start posting all and everything just to keep up with thé ADS and stuff , anyway Always glad to see your efforts in sharing quality stuff with thé community
@@miss_tech ads? 🤔 I don't think I understand what I've done to disappoint you. I've been covering AI extensively in the past few weeks and it seem like you enjoy this subject... so I don't know why you said "long time". If you enjoy web scraping - that's great! But I have so many tutorials on it already, it seems unnescessary to repeat the same actions just on different sites... I always post polls asking viewers what they want to see next and which series they like more - but it seems that most people prefer other subjects (like GUI and AI). I'm sorry to hear you don't like any of my recent tutorials, but I'm still not sure why... especially how it had to do with ads. My inbox is full of sponsor emails which I never replied to. I've filmed a single sponsored video with Tabnine a very long time ago - I did it once and I don't do it since. What makes you think I'm showing you ads?
Well I am late sorry about that 😄 you just over the live stream but I'll watch it again and thanks for the video I will definitely send it to my family I hope they will like your video 👌😄 but they didn't understand because they don't know how to do programming😂😂 thumbnail is really good 😂👌
hahahaha that's awesome Ubdhav!! My mom sent a few of my videos to my Grandmother and she absolutely LOVED them! Not only she's not a programmer, but she also doesn't even speak English!!! hahahaha 🤣🤣🤣 she's obviously not bias! 😂 Thank you for all the support! If you watched the previous ML tutorials - you gonna really like this one! it connects all the dots together and in the end you'll have your very own AI project to experiment with 😉
@@PythonSimplified thanks youu Maria because you helped I mean your videos helped me so much and my grandmother is not alive if she alive she definitely LOVE your video and yeah same she doesn't know English too ✌️😄have a nice day. and one more think please correct my name it's (udbhav)😂👌
Greetings from Colombia. We look gradient decent in action, but not the target predicted (y^) in terms of zero or one. I think we have to add a step function after the sigmoid function. Without a binary prediction we cannot use the algoritm to solve the problem we wanted. How to use the predicted value? Have to we use any value as a boundary? for instance: 1 if y^>=0.95 else 0. Thanks for video
Marina, Perhaps you have no idea about What you are giving…. Thank you. Be blessed in all your entire live. You are loosing anything. Your vids are great moments to meet you and to learn clearly from you. I have problem to create the generate_data() function - as you did it - with my new MacBook Pro Big Sur.
Thank you so much Prajwal! 😁 I was planning to do some encryption/decryption tutorials in the near future... we'll probably start with decrypting the cypher from Edgar Allan Poe's "The Gold-Bug". I'm not a big expert in digital signature algorithms yet, but if I see lots of requests I'll look into it 😉 Thanks for requesting!
@@PythonSimplified Hey Mariya hope U R doing well actually I meant was DATA STRUCTURES AND ALGORITHMS by DSA in python my bad 😂😜 I am 2nd year of my engineering and we have this really important topic of DATA STRUCTURES AND ALGO. And SURE I UNDERSTAND NO PROBLEM👍 I look forward to your crypting the cypher from Edgar Allan Poe's "The Gold-Bug". video keep going 😁🖖
Great video Mariya. Any tuto about recursion in python? I need to know how The Towers of Hanoi works. I am a beginner 😊. Love your tutorials and smile. Keep going the best
Aha, Towers of Hanoi! someone is doing a degree in Computer Science, eh? 😉 Recursion actually sounds like a great idea! I'll definitely post a video on it shortly, not sure if a concept video or code-along, but I'll add it to my "to do" list 😀😀😀
I heard about you on the Superdatascience podacast. Do you realize how incredible you are? You saved me a lot through my Deep Learning course with this series
great video as usual, and hope to have actual object for training, like in game playing or ocr prediction, and I was wondering at the end of the video, when you said, we need to work with the learning rate , epoch, etc. I though that , could we make training model for these values that work with the main training model until we get perfect state ?!! , Thank you Maria ❤❤
Thank you so much dear!! 😁😁😁 And absolutely! you would usually automate the process of trying different values for epochs, learning rate, etc (and later on we'll go over hidden dimensions, different layers with different activation functions like ReLU and all kinds of other fun stuff). Since there are so many variables and each database requires different approaches - we set a few initial values rather then one and evaluate the different results on smaller bathes of the data, rather than the entire database. We'll get to the best techniques for that in future episodes 😉
@@PythonSimplified great, I will follow till the end..keep forward, it was little tough at the beginning, but I understand the concept, when I remembered some other cases that was having hidden code of model learning, but the object was like trying to teach a case of game player to win, or 3d object to move and walk like human, and in all cases score was the trophy, and hence it was keep training, I saw many videos about that, but no code !!, now I saw the code and linked the final score of many features , targets, values to these cases, so, we need to see your code in action on tangent cases , I hope that I hit the point, and am not wrong in my deduction
Thank you so much Alan! 😊 Voice recognition is an incredible project! I'm still grasping how sound data is represented in numbers... the entire concept of amplitude/frequency is very interesting and I really wish I understood it better 🙂 When you finish your program - definitely give me a shout, I would love to see what you came up with! 😀
Hi Sofia! my apologies!! the link was broken, so I've loaded everything to my Github and will figure out how to store it again on Wayscript shortly 😉 You can find it here: github.com/MariyaSha/TrainBasicNN
Hello, nice video. I'm confuse about your function generated data. May be the function should written like : generate_data(n_values, n_features). In my understanding, the features are columns name and the values (n_samples) represented by rows. So, if you write generate_data(4,3) it will return 4 rows (n_values or n_samples) and 3 columns (n_features --> x0, x1, x2) of data, May be it's my mistake in understanding, would you enlighten me?
Hi Yunfei, thank you! 😃 A feature in the context of ML has a different meaning, checkout my intro to ML video for more details 😉: ua-cam.com/video/mMc_PIemSnU/v-deo.html
Yup, this video is definitely on the long side! 😅 But it's definitely worth it! it breaks down the entire training with gradient descent algorithm into handy functions... annnnnd we get an opportunity to practice our supreme Numpy skills 💪😎
Hi Maria, I am a little confused! If you have 4 features and 3 values for each feature, then shouldn't you have 3 rows of data and 4 columns with each column representing a separate feature column as x1,x2, x3, and x4 (or x0,x1,x2,x3 if you like)?
Just a detail... "pd" (when you imported pandas) is a is like the n-word... But to talk about gay people in France 😅 (why do we have so much different bad words in France ? 😑) (Thank you for all the content you make, it's always easy to understand and it helps a lot)
hahahahaha this is hilarious! 🤣🤣🤣 The French "pd" is actually similar to the Russian "pedik" which has the exact same meaning hahahaha I bet you the guys at Pandas were neither French, nor Russian... or maybe they were just really funny people hahaha 😅 (and thank you so much for the beautiful feedback! so happy to hear that dear! 😁)
Thank you so much Venkat! glad you like my tutorials! 😀 I live in British Columbia, Canada. Originally from Crimea and grew up in the Middle East (I'm basically from everywhere 🤣🤣🤣)
hi ! very nice ! i have a question though: what happens by ranndom you get the prediction to be 0 or 1 ? in the cross ntropy you get log10(0) wich is nan...
Hi Ahmed 😀 Gradient descent requires a floating point input between 0 and 1, while the sigmoid function provides exactly that as its output - so they're a perfect match! If we were to use a different activations function, such as the step function we've used in the perceptron tutorial - we wouldn't be able to combine it with gradient descent. I talk about it on 00:37 in my gradient descent video: ua-cam.com/video/jwStsp8JUPU/v-deo.html You can also try entering 0 or 1 in the gradient descent formula instead of the floating point prediction value - you'll see that it's not gonna give you a valid result. Hope it helps! 😊
@Ahmed El sorry I don't know if I understood your question... I'll try to answer anyways 😅 hahaha Sigmoid is basically used to calculate probability. It can also be used for classification purposes as if the probability is higher than a certain threshold (bias) - you can classify the input as 1. If it's below a certain threshold - you can classify it as 0. This will be done during the inference stage (also known the prediction stage) which is not covered by this tutorial. This is usually done with brand new data, that the model wasn't exposed to during the training process. In my personal opinion, probabilities are much better for classification than working with binary output. As with probabilities you get to see exactly what chances are for the input to belong to a certain class. In other words: "this is a cat" is very general. But "this is 70% likely to be a cat" is a bit more specific and the key note here: If something is only "51% likely to be a cat" - you'll never know how uncertain your model is if you're using a binary step function 😉 I hope I understood your question, sorry if I answered about something else 🙃 Please let me know if I got you right!
Hi guys, sorry about the mess-up!! the link got broken a few months back and I only noticed now 😭😭😭 I've just loaded everything to my Github, you can find it here: github.com/MariyaSha/TrainBasicNN
I explain it in the very beginning of the video, check out the intro 🙂 It's focusing on the mathematical operations rather than on the shortcuts, it helps you understand how the algorithms work behind the scenes.
It's a tough one, as I'm yet to see even one source that teaches these subjects clearly... 🤔 I started learning from Udacity, but looking back in my old notebooks - I see that I didn't understand a thing! The only way I was able to learn is through lots of online research, reading articles and experimenting with code. It's very weird to say that - but I believe you already found the best source to learn from on my channel... I very rarely refer to my own videos as "the best", but if someone would explain ML to me the same way I've explained it to myself- it would save me 2 years of pondering! It's an exciting journey, but it's hard to find clear information... that's why I had to start this series! 😊
your videos are great and im always looking through your material before anyone else if i wanted to learn python, but woman, youre so beautiful i cant even
סקאם?? גדלתי בקריית שמונה... ולפי התמונות באינסטרגם אני לא בדיוק מסתירה את זה 😉 לא הייתי בארץ המון שנים, אבל המשפחה שלי עדיין גרה שם 😊 זה לא נוגד את זה שנולדתי באוקראינה ואני חיה בקנדה, אני פשוט מהגרת סדרתית חחחחחח
I need to learn Spanish to do that! 😅 hahaha I do understand quite a lot, I've been watching Argentinian soap operas all my childhood so much of the language is subcountiously embedded in me... but in terms of speaking, I'm more in the "Donde Esta La Biblioteca" level 🤣🤣🤣 hahahahaha
Good video again !!
I tell you that through your web scraping video with selenium I was able to program a bot to use it in my work test and achieve the job position thanks to that.
Thanks again for teaching without any interest and without knowing that you can help people achieve their goals.
oh wow!!! I'm so happy to hear that!!! 😁😁😁
Great job on the the bot and congrats on the new position!!! I'm super happy I could help and thank you so much for your beautiful comment, Martin! 😊
Great video as always 😊 and the Thumbnail is HILARIOUS 🤣
Thank you so much Debayan, super happy you liked it! 😀
I had the funniest photoshoot in the world when filming for the thumbnail - I have no megaphone nor a sporty headband... so if you look at all the original photos, you'll see an Axl Rose impersonator holding a makeup remover and shouting on imaginary creatures 🤣🤣🤣 the power of graphic design is only limited by ones imagination! hahaha
Thank you for this video! That’s the kind of coding level breakdown we need to understand how machine learning is actually working! ♥️♥️♥️
Thank you so much for the lovely comment!!! Super happy you you enjoyed this tutorial! 😁😁😁
Your a better that most of the paid instructors I have had. IKeep doing the people work! We appreciate you!
thank you so much and learned a lot from your talent. greetings from Ethiopia🇪🇹🇪🇹
Hey maria, I can't believe how incredibly awesome it was. I have never seen in my life someone train a ML model only using numpy and pandas. It's incredible, hats off!! Love from India! ♥️♥️
Masha I am quite new in this field and you are a super clear teacher. Following you it is a pleasure. A big Thank. I just wanted to get you a beer for your kindness. Ciao (I am italian!)
For those of you who got confused about why the "np.dot(feature, weights)" worked on two matrix of the dimentions 3X1 since it supposed to work as matrix multiplication. I found out that, the "np.dot(feature, weights)" will convert the second one dimentional array which is in this case the array "weights" to a column matrix. In other words, the dimention of the matrix "weights" will be converted from 1X3 to 3X1
Thank you so much for the awesome explanation!! 😃😃😃
This series has been so amazing. You are a great teacher! You have given me more confidence for my upcoming internship. Thank you for doing this!
Thank you so much for the incredible comment!!! Super happy I was able to help on your exciting journey!! 😃😃😃
Mariya, show these AIs NO MERCY!!! 🤣
hahahahahahaha I shall not!!! they will pay for what they did in The Terminator!!! 🤣🤣🤣
They better lose those pounds before I pound them!!! 💪 hahahaha
@@PythonSimplified HELL YEAH!!! We survived August 29 1997 and we shall not let 2029 happen EVER!!! Now, when you say "lose pounds", can that mean "junk data"??? 🤣🤣🤣 🥁Siempre un gusto ver tu lecciones bella Mariya ❤ Gracias
You are the best who explain machine learning. All vids are short and simple, covering everything quickly, and conveying the idea in a clear way.
I appreciate your hard work on the thumbnails and the content is also on high level, well done
Thank you so much for the lovely feedback, Jim! glad you liked it! 😁
Thanks! I tried doing this on paper during my PhD but it is just so laborious that I got bored lol! But now with Python it is so easy to run through the basic operation of the perceptron. Awesome!
Love from Nigeria. Keep up the good work Mariah
Thank you so much Cyberxploit Hausa!! 😁
Cheers from Canada! 😀🍁
Well done, I really like the way you describe the topic. You do it like playing a nice melody, actually a symphony.
Precious video thanks a lot Maria ❤
Thanks for share your knowledge with us ❤
I love the thumbnail hahaha. Amazing video my friend, I wasn't able to come to the premiere but I hope next time I'll be there, have a wonderful day :).
Thank you so much! Glad you liked it!! 😁😁😁
We'll catch up next time, don't worry about it! 😉
Really really thank you for this video!! First of all, your natural pedagogic system is amazing! Your english is clear, so the people like me who live in Argentina can understand all and practice the lenguage. About the step by step code, is great, i can see the importance of having a CLEAN CODE! With order, details functions, so you can see it like a dron camera! jaja This world of AI is amazing... now, please, give us some time to practice and to naturalice the concepts of these lasts videos :) would be the next one an app with AI? mmm , thanks again!
Thank you so much for the incredible feedback!!! 😊
I'm actually about to get back to this AI series from a more practical perspective 😉 We will begin diving into Pytorch and ultimately YES - even combining AI in your personal apps 😁😁😁
love you
please don't ever think for even a second to quit publishing amazing videos
Thank you so much Ehsan! 😁
Don't worry, I'm not going anywhere anytime soon, this is just the beginning! 😉
Brief correction: in cross entropy we use natural logs, instead of log10.
Because of the activation of the network, this normalises the prediction. E.g. sigmoid and softmax use E^x, and to normalise an exp, because exp is the opposite of natural logs, we use log, not log10 since it’s not a 10^x problem.
I will definitely re-listen all of this. Thank you for your patience and thoroughness ( povtorenie - mat' uchenija !)
Parabéns pelo excelente trabalho.
Great video 😍😍 also the thumbnail is really good🤩🤩
Thank you so much Fire Ball! Glad you liked it! 😁😁😁
Thank you for making this channel, I’m new to coding and this channel is the best
Thank you so much dear! I'm so happy to hear that! 😀
Brilliantly done, Mariya. Do you have a theoretical reference for why the average_loss always converges to around 0.27 regardless of the number of weights or the number of training datasets? Random x's and random 0/1's is a stastically stationary system, i think, so the average_loss should always be about the same. Do you have a reference for calculating the average loss with a formula rather than by simulation?
Nice
Thank you! :)
I do love the way you explain everything, thanks.
Nice cover ❤
Thank you so much! glad you liked it! 😀
wow you amazing
vey very informative
You're just AMAZING. Your explanations are so smart you could make a 5yo kid understand machine learning!
Awesome presentation. And after a journey across hundreds of videos, describing this, this video had me 💡have a light 💡 bulb go off.
Haha That thumbnail! I'm watching this later, don't know absolutely anything about "Neural" stuff and want to see if I can learn something! Thanks!
Hi Lopo! 😀
Before you watch this one, make sure you've seen the previous episodes where I explain all the neural stuff in detail 😉:
Episode 1 - Perceptron:
ua-cam.com/video/-KLnurhX-Pg/v-deo.html
Episode 2 - Cross Entropy Loss:
ua-cam.com/video/EJRFP3WmS6Q/v-deo.html
Episode 3 - Gradient Descent:
ua-cam.com/video/jwStsp8JUPU/v-deo.html
This tutorial connects all the dots together and in the end you'll have your very own AI project to experiment with! 😁
@@PythonSimplified Oh! I will watch them first, thank you very much!!!
Excellent clarity !!!
U nailed it...
Very Nice.
Nice job
excellent video,you can't imagine how much you helped me
Yeeey!! Thank you so much Alan, I'm super happy to help! 😊
Great tutorial, keep em coming
Thank you so much Kay! 😁 I sure will!
finally a quality conent video since long time ;)
Thank you!... since long time? 🙃
@@PythonSimplified yes since then , from m'y perspective i Saw that ur content was awsome when u started your Channel step by step with that python and Scraping intros, but from months ago u start posting all and everything just to keep up with thé ADS and stuff , anyway Always glad to see your efforts in sharing quality stuff with thé community
@@miss_tech ads? 🤔 I don't think I understand what I've done to disappoint you. I've been covering AI extensively in the past few weeks and it seem like you enjoy this subject... so I don't know why you said "long time".
If you enjoy web scraping - that's great! But I have so many tutorials on it already, it seems unnescessary to repeat the same actions just on different sites...
I always post polls asking viewers what they want to see next and which series they like more - but it seems that most people prefer other subjects (like GUI and AI). I'm sorry to hear you don't like any of my recent tutorials, but I'm still not sure why... especially how it had to do with ads. My inbox is full of sponsor emails which I never replied to.
I've filmed a single sponsored video with Tabnine a very long time ago - I did it once and I don't do it since.
What makes you think I'm showing you ads?
Import Mariya as youtube ...You Make My Life Easy ...thank you so much :)
I suggest to import me as "yt" instead - it will be easier to code 😆 hahaha thank you so much! 😁😁😁
Super happy to hear that!
Right now AI is one of my subject in my studying.... 😀.... Thank you simplified for giving me a revision feel ..
That's awesome Jay! what a great timing for an AI series, eh? 😁
@@PythonSimplified yeah! 😀
you are awsome teaching, i realize i need a lot of learning
Well I am late sorry about that 😄 you just over the live stream but I'll watch it again and thanks for the video I will definitely send it to my family I hope they will like your video 👌😄 but they didn't understand because they don't know how to do programming😂😂 thumbnail is really good 😂👌
hahahaha that's awesome Ubdhav!! My mom sent a few of my videos to my Grandmother and she absolutely LOVED them! Not only she's not a programmer, but she also doesn't even speak English!!! hahahaha 🤣🤣🤣 she's obviously not bias! 😂
Thank you for all the support! If you watched the previous ML tutorials - you gonna really like this one! it connects all the dots together and in the end you'll have your very own AI project to experiment with 😉
@@PythonSimplified thanks youu Maria because you helped I mean your videos helped me so much and my grandmother is not alive if she alive she definitely LOVE your video and yeah same she doesn't know English too ✌️😄have a nice day. and one more think please correct my name it's (udbhav)😂👌
I love your class.thank you so much
Thank you Roberto! happy to hear that! 😀
Really great example. Thanks.
You are a great teacher. Thank you!🙏
Greetings from Colombia. We look gradient decent in action, but not the target predicted (y^) in terms of zero or one. I think we have to add a step function after the sigmoid function. Without a binary prediction we cannot use the algoritm to solve the problem we wanted. How to use the predicted value? Have to we use any value as a boundary? for instance: 1 if y^>=0.95 else 0. Thanks for video
Thank you very much
Simply awesome!!!!!
Simply thank you so much!!! 😀😀😀
You are simply the best!
Thank you so much! 😊
Marina,
Perhaps you have no idea about What you are giving…. Thank you. Be blessed in all your entire live. You are loosing anything. Your vids are great moments to meet you and to learn clearly from you.
I have problem to create the generate_data() function - as you did it - with my new MacBook Pro Big Sur.
The thumbnails are getting better everyday 😂
Thank you so much Guilherme! I really liked this one too 😁😁😁
very nice
great work
Thank you! Glad you liked it 😊
Awesome thumbnail Mariya
Thank you so much GUJJJEK!! 😀
That's the first thing that comes to my mind when I hear "train a neural network" hahahaha 🤣🤣🤣
So great, thx
Great material, after your explanation I understood step by step how it works :-)
Great tutorial Mariya keep going !!!!😄 I love your tutorials,
hey if possible could also create tutorials on DSA in Python
Thank you so much Prajwal! 😁
I was planning to do some encryption/decryption tutorials in the near future... we'll probably start with decrypting the cypher from Edgar Allan Poe's "The Gold-Bug". I'm not a big expert in digital signature algorithms yet, but if I see lots of requests I'll look into it 😉 Thanks for requesting!
@@PythonSimplified Hey Mariya hope U R doing well actually I meant was
DATA STRUCTURES AND ALGORITHMS
by DSA in python my bad 😂😜
I am 2nd year of my engineering and we have this really important topic of DATA STRUCTURES AND ALGO.
And SURE I UNDERSTAND NO PROBLEM👍
I look forward to your crypting the cypher from Edgar Allan Poe's "The Gold-Bug".
video
keep going 😁🖖
Fabulous as usual 😉
Great video Mariya. Any tuto about recursion in python? I need to know how The Towers of Hanoi works. I am a beginner 😊. Love your tutorials and smile. Keep going the best
Aha, Towers of Hanoi! someone is doing a degree in Computer Science, eh? 😉
Recursion actually sounds like a great idea! I'll definitely post a video on it shortly, not sure if a concept video or code-along, but I'll add it to my "to do" list 😀😀😀
Where did you calculate the derivatives?
Grazie.
Спасибо!
На здоровье! рада помочь 😀
I heard about you on the Superdatascience podacast.
Do you realize how incredible you are? You saved me a lot through my Deep Learning course with this series
great video as usual, and hope to have actual object for training, like in game playing or ocr prediction, and I was wondering at the end of the video, when you said, we need to work with the learning rate , epoch, etc. I though that , could we make training model for these values that work with the main training model until we get perfect state ?!! , Thank you Maria ❤❤
Thank you so much dear!! 😁😁😁
And absolutely! you would usually automate the process of trying different values for epochs, learning rate, etc (and later on we'll go over hidden dimensions, different layers with different activation functions like ReLU and all kinds of other fun stuff). Since there are so many variables and each database requires different approaches - we set a few initial values rather then one and evaluate the different results on smaller bathes of the data, rather than the entire database. We'll get to the best techniques for that in future episodes 😉
@@PythonSimplified great, I will follow till the end..keep forward, it was little tough at the beginning, but I understand the concept, when I remembered some other cases that was having hidden code of model learning, but the object was like trying to teach a case of game player to win, or 3d object to move and walk like human, and in all cases score was the trophy, and hence it was keep training, I saw many videos about that, but no code !!, now I saw the code and linked the final score of many features , targets, values to these cases, so, we need to see your code in action on tangent cases , I hope that I hit the point, and am not wrong in my deduction
Ххахаха nice preview)
Спасибо Андрей! 😁
This is fascinating. I am building a voice recognition / ai intent program that works offline
Thank you so much Alan! 😊
Voice recognition is an incredible project! I'm still grasping how sound data is represented in numbers... the entire concept of amplitude/frequency is very interesting and I really wish I understood it better 🙂
When you finish your program - definitely give me a shout, I would love to see what you came up with! 😀
@@PythonSimplified I sure will, thanks
where can i find the code? Because the links don't work. Thank you for the help!
Hi Sofia! my apologies!! the link was broken, so I've loaded everything to my Github and will figure out how to store it again on Wayscript shortly 😉
You can find it here: github.com/MariyaSha/TrainBasicNN
Great tutorial !!!
)) Your eyes' color is green 😉
How old are you miss if not secret miss ?
Hello, nice video. I'm confuse about your function generated data. May be the function should written like : generate_data(n_values, n_features). In my understanding, the features are columns name and the values (n_samples) represented by rows. So, if you write generate_data(4,3) it will return 4 rows (n_values or n_samples) and 3 columns (n_features --> x0, x1, x2) of data, May be it's my mistake in understanding, would you enlighten me?
Great work. I thought a feature is a column and not a row in a data set. It looks like you call a row as a feature in a data set?
Hi Yunfei, thank you! 😃
A feature in the context of ML has a different meaning, checkout my intro to ML video for more details 😉:
ua-cam.com/video/mMc_PIemSnU/v-deo.html
Add a visual assistance of the same maybe in terms of layers and functions including inputs and outputs
any advice on scrapping contact and name list of person for advertising purposes lets say example whitepages ..
I need a bigger cup to hold more tea.
Yup, this video is definitely on the long side! 😅
But it's definitely worth it! it breaks down the entire training with gradient descent algorithm into handy functions... annnnnd we get an opportunity to practice our supreme Numpy skills 💪😎
Hi Maria, I am a little confused! If you have 4 features and 3 values for each feature, then shouldn't you have 3 rows of data and 4 columns with each column representing a separate feature column as x1,x2, x3, and x4 (or x0,x1,x2,x3 if you like)?
Just a detail... "pd" (when you imported pandas) is a is like the n-word... But to talk about gay people in France 😅 (why do we have so much different bad words in France ? 😑)
(Thank you for all the content you make, it's always easy to understand and it helps a lot)
hahahahaha this is hilarious! 🤣🤣🤣
The French "pd" is actually similar to the Russian "pedik" which has the exact same meaning hahahaha
I bet you the guys at Pandas were neither French, nor Russian... or maybe they were just really funny people hahaha 😅
(and thank you so much for the beautiful feedback! so happy to hear that dear! 😁)
Love you mariya ❤
Your videos are just awesome
You always clear my doubts
Thank you🙏🙏🙏
Can I know where are you from?
Thank you so much Venkat! glad you like my tutorials! 😀
I live in British Columbia, Canada. Originally from Crimea and grew up in the Middle East (I'm basically from everywhere 🤣🤣🤣)
@@PythonSimplified Maria, can you upload some tutorial video about pygame ?
hi ! very nice ! i have a question though: what happens by ranndom you get the prediction to be 0 or 1 ? in the cross ntropy you get log10(0) wich is nan...
nevermind lol ive messed up some stuff ... lol
well done, girl =)
Thank you! 😊
Wow
if the Segmoid function is the activation here shouldn't it return 0 or 1 why you are just returning its output directly as the prediction ?
Hi Ahmed 😀
Gradient descent requires a floating point input between 0 and 1, while the sigmoid function provides exactly that as its output - so they're a perfect match!
If we were to use a different activations function, such as the step function we've used in the perceptron tutorial - we wouldn't be able to combine it with gradient descent.
I talk about it on 00:37 in my gradient descent video: ua-cam.com/video/jwStsp8JUPU/v-deo.html
You can also try entering 0 or 1 in the gradient descent formula instead of the floating point prediction value - you'll see that it's not gonna give you a valid result.
Hope it helps! 😊
@@PythonSimplified thanks, but isn't this algorithm using Segmoid supposed to be classification logistic regression ?
@Ahmed El sorry I don't know if I understood your question... I'll try to answer anyways 😅 hahaha
Sigmoid is basically used to calculate probability. It can also be used for classification purposes as if the probability is higher than a certain threshold (bias) - you can classify the input as 1.
If it's below a certain threshold - you can classify it as 0.
This will be done during the inference stage (also known the prediction stage) which is not covered by this tutorial. This is usually done with brand new data, that the model wasn't exposed to during the training process.
In my personal opinion, probabilities are much better for classification than working with binary output. As with probabilities you get to see exactly what chances are for the input to belong to a certain class.
In other words: "this is a cat" is very general. But "this is 70% likely to be a cat" is a bit more specific and the key note here:
If something is only "51% likely to be a cat" - you'll never know how uncertain your model is if you're using a binary step function 😉
I hope I understood your question, sorry if I answered about something else 🙃
Please let me know if I got you right!
We are trying to find patterns in randomness
The page with the code can't be found :( Is there an updated ling anywhere?
@PythonSimplified, do you have an updated link. It didn’t work
Hi guys, sorry about the mess-up!! the link got broken a few months back and I only noticed now 😭😭😭
I've just loaded everything to my Github, you can find it here: github.com/MariyaSha/TrainBasicNN
@@PythonSimplified 👍
I love learn python because of you gorgeous
why you didnt used here keras and tensorflow lib.
I explain it in the very beginning of the video, check out the intro 🙂
It's focusing on the mathematical operations rather than on the shortcuts, it helps you understand how the algorithms work behind the scenes.
What the monitor model ??
Hii, I'm interested to study ML, DL, AI but where can I start first ?
It's a tough one, as I'm yet to see even one source that teaches these subjects clearly... 🤔
I started learning from Udacity, but looking back in my old notebooks - I see that I didn't understand a thing! The only way I was able to learn is through lots of online research, reading articles and experimenting with code.
It's very weird to say that - but I believe you already found the best source to learn from on my channel... I very rarely refer to my own videos as "the best", but if someone would explain ML to me the same way I've explained it to myself- it would save me 2 years of pondering!
It's an exciting journey, but it's hard to find clear information... that's why I had to start this series! 😊
@@PythonSimplified wow! Thank you
your videos are great and im always looking through your material before anyone else if i wanted to learn python, but woman, youre so beautiful i cant even
Hello Mariya, thanks for the tutorial. I need a course in this topic. I want to build a model in Numpy. Can you help?
yes it do loss it decreasing so how use it. gice example. best is lottery numbers lol
I watch your videos because of your eyes 😂👌
hahahaha thank you! so I can switch to make up tutorials and you'll still be onboard?? 🤣🤣
@@PythonSimplified doesn't matter I will always follow you : ) lol
Can you show us how to create a AI program to win Draft Kings? Cuz for science. Nah i just want AI to win bets... FOR SCIENCE....
Lovely code by lovely girl
Discord server link???
Here you go, I posted it in the livechat 😉
discord.gg/GbgZaCMA6A
How to test
שלום, חשבתי את בכלל מאוקראינה, גיליתי קנדה אז הלכתי לאינסטגרם. מה היית חיילת שיאו, מה זה הסקאם הזה
סקאם?? גדלתי בקריית שמונה... ולפי התמונות באינסטרגם אני לא בדיוק מסתירה את זה 😉
לא הייתי בארץ המון שנים, אבל המשפחה שלי עדיין גרה שם 😊
זה לא נוגד את זה שנולדתי באוקראינה ואני חיה בקנדה, אני פשוט מהגרת סדרתית חחחחחח
Please do tutorials in spanish 😊❤
I need to learn Spanish to do that! 😅 hahaha
I do understand quite a lot, I've been watching Argentinian soap operas all my childhood so much of the language is subcountiously embedded in me... but in terms of speaking, I'm more in the "Donde Esta La Biblioteca" level 🤣🤣🤣 hahahahaha
Anyone else who is in love?
One Question: Do you use eye lense? 🤓
0 dislikes lets goo
just give it an hour or two...🤪
I always have the occasional haters tagging along, sometimes even before the video premieres 😅 hahahaha
Lol
how old are u? :) hahaha when i grow up i want to be like you, im 22 lol
hahahaha did you just assume I'm older than 22, Lucas??? 🤣🤣🤣
I actually am, not by much, but a real lady never tells her true age! 🤪