title

Neural Networks from Scratch - P.1 Intro and Neuron Code

description

Building neural networks from scratch in Python introduction.
Neural Networks from Scratch book: https://nnfs.io
Playlist for this series: https://www.youtube.com/playlist?list=PLQVvvaa0QuDcjD5BAw2DxE6OF2tius3V3
Python 3 basics: https://pythonprogramming.net/introduction-learn-python-3-tutorials/
Intermediate Python (w/ OOP): https://pythonprogramming.net/introduction-intermediate-python-tutorial/
Mug link for fellow mug aficionados: https://amzn.to/2xcyfPC
Channel membership: https://www.youtube.com/channel/UCfzlCWGWYyIQ0aLC5w48gBQ/join
Discord: https://discord.gg/sentdex
Support the content: https://pythonprogramming.net/support-donate/
Twitter: https://twitter.com/sentdex
Instagram: https://instagram.com/sentdex
Facebook: https://www.facebook.com/pythonprogramming.net/
Twitch: https://www.twitch.tv/sentdex
#nnfs #python #neuralnetworks

detail

{'title': 'Neural Networks from Scratch - P.1 Intro and Neuron Code', 'heatmap': [{'end': 200.682, 'start': 172.56, 'weight': 1}], 'summary': 'Series covers creating neural networks from scratch in python using numpy, discussing challenges in image classification, emphasizing simplicity in mathematical functions, outlining prerequisites, and explaining the structure and function of neural networks while coding the first neuron and understanding neurons in neural networks.', 'chapters': [{'end': 249.93, 'segs': [{'end': 54.024, 'src': 'embed', 'start': 25.169, 'weight': 3, 'content': [{'end': 30.632, 'text': "we're going to show first in truly raw Python, no libraries or no third-party libraries.", 'start': 25.169, 'duration': 5.463}, {'end': 33.994, 'text': "And then we're going to use NumPy for multiple reasons.", 'start': 31.252, 'duration': 2.742}, {'end': 35.595, 'text': 'NumPy just makes a lot of sense here.', 'start': 34.074, 'duration': 1.521}, {'end': 42.758, 'text': "It's an extremely useful library and it'll cut our lines of our full application down a ton.", 'start': 35.675, 'duration': 7.083}, {'end': 43.959, 'text': "It'll make it much faster.", 'start': 42.898, 'duration': 1.061}, {'end': 45.56, 'text': 'And NumPy is a great thing to learn.', 'start': 44.439, 'duration': 1.121}, {'end': 49.462, 'text': "So we will show everything from scratch in Python first, and then we're going to use NumPy.", 'start': 45.62, 'duration': 3.842}, {'end': 54.024, 'text': 'Now, why would anybody want to do this to themselves? All right.', 'start': 49.622, 'duration': 4.402}], 'summary': 'Comparing raw python vs numpy for efficiency and speed in code.', 'duration': 28.855, 'max_score': 25.169, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI25169.jpg'}, {'end': 96.179, 'src': 'embed', 'start': 66.417, 'weight': 1, 'content': [{'end': 71.101, 'text': 'The purpose is to actually learn how neural networks work at a very deep level,', 'start': 66.417, 'duration': 4.684}, {'end': 76.886, 'text': 'so that when we go back to whatever framework we actually use be it PyTorch or TensorFlow,', 'start': 71.101, 'duration': 5.785}, {'end': 82.571, 'text': "Keras or maybe some library that doesn't even exist yet we actually understand what we're doing.", 'start': 76.886, 'duration': 5.685}, {'end': 86.454, 'text': 'So, for myself at least, when I learned deep learning, it was all, you know.', 'start': 82.671, 'duration': 3.783}, {'end': 91.417, 'text': 'yes, it was hard, but at the same time everything was kind of solved for me.', 'start': 86.454, 'duration': 4.963}, {'end': 96.179, 'text': 'you know how many layers, how many nodes per layer, what activation functions to use, like all that stuff.', 'start': 91.417, 'duration': 4.762}], 'summary': 'Deeply understanding neural networks for proficient usage in frameworks like pytorch or tensorflow.', 'duration': 29.762, 'max_score': 66.417, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI66417.jpg'}, {'end': 147.673, 'src': 'embed', 'start': 108.443, 'weight': 4, 'content': [{'end': 114.524, 'text': 'And this became a problem when I tried to solve problems that had not yet been solved for me.', 'start': 108.443, 'duration': 6.081}, {'end': 119.786, 'text': 'So classifying handwritten images of handwritten digits, pretty darn simple.', 'start': 115.065, 'duration': 4.721}, {'end': 125.655, 'text': 'Taking that a step further and classifying images of cats and dogs, Pretty simple.', 'start': 120.285, 'duration': 5.37}, {'end': 138.245, 'text': 'But then taking that just a tiny step further and classifying images that were instead frames from a video game and trying to map that to actions that I want to take in the video game,', 'start': 126.876, 'duration': 11.369}, {'end': 139.846, 'text': "suddenly I'm lost.", 'start': 138.245, 'duration': 1.601}, {'end': 142.549, 'text': "And that's no good.", 'start': 140.247, 'duration': 2.302}, {'end': 145.551, 'text': "And there's no way to really know where to go next.", 'start': 142.609, 'duration': 2.942}, {'end': 147.673, 'text': "Like you can see that there's a problem,", 'start': 145.591, 'duration': 2.082}], 'summary': 'Challenges in solving image classification problems, progressing from digits to video game frames.', 'duration': 39.23, 'max_score': 108.443, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI108443.jpg'}, {'end': 208.264, 'src': 'heatmap', 'start': 172.56, 'weight': 0, 'content': [{'end': 180.568, 'text': "you've got your input data and every unique input and every unique neuron The information coming through has a unique weight associated with it.", 'start': 172.56, 'duration': 8.008}, {'end': 184.211, 'text': 'Those get summed together per neuron plus a bias.', 'start': 180.728, 'duration': 3.483}, {'end': 192.72, 'text': 'run through an activation function, and then we do that for every single layer, giving us the output information.', 'start': 184.838, 'duration': 7.882}, {'end': 200.682, 'text': 'from there we want to calculate a loss, which is a calculation of how wrong the neural network is, so that hopefully we can fix it.', 'start': 192.72, 'duration': 7.962}, {'end': 208.264, 'text': 'and at the end of that, even though that was just the forward pass, that looks already extremely, extremely daunting.', 'start': 200.682, 'duration': 7.582}], 'summary': 'Neural network processes input data, assigns unique weights, calculates loss for improvement.', 'duration': 39.886, 'max_score': 172.56, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI172560.jpg'}, {'end': 256.591, 'src': 'embed', 'start': 227.88, 'weight': 8, 'content': [{'end': 230.682, 'text': 'Rectified linear is actually just the max of zero or the output.', 'start': 227.88, 'duration': 2.802}, {'end': 232.263, 'text': 'We do this for all of the layers,', 'start': 230.922, 'duration': 1.341}, {'end': 241.407, 'text': 'then we have our softmax activation function at the very end and then the cross entropy losses actually turns out to be a negative log loss.', 'start': 232.263, 'duration': 9.144}, {'end': 244.328, 'text': 'so just due to the nature of neural networks.', 'start': 241.407, 'duration': 2.921}, {'end': 249.93, 'text': 'so this is the entire forward pass and calculation of loss formula.', 'start': 244.328, 'duration': 5.602}, {'end': 256.591, 'text': "i encourage you to look at all of the functions and things that we're doing here to determine if any of this is really over your head,", 'start': 249.93, 'duration': 6.661}], 'summary': 'Neural network uses rectified linear and softmax activation functions, with cross entropy loss. forward pass and loss calculation explained.', 'duration': 28.711, 'max_score': 227.88, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI227880.jpg'}], 'start': 5.425, 'title': 'Neural networks and image classification challenges', 'summary': 'Covers creating neural networks from scratch in python using numpy, emphasizing self-learning for understanding frameworks. it also discusses challenges in image classification, from simple tasks to complex ones, highlighting sudden complexity and lack of direction.', 'chapters': [{'end': 107.582, 'start': 5.425, 'title': 'Neural networks from scratch', 'summary': 'Covers creating neural networks from scratch in python, with a focus on understanding the deep level workings of neural networks, using numpy for efficiency, and the importance of self-learning for a deeper understanding of frameworks like pytorch or tensorflow.', 'duration': 102.157, 'highlights': ['The chapter emphasizes creating neural networks from scratch in Python to deeply understand how neural networks work, rather than just programming a framework, with the goal of gaining a deeper understanding of frameworks like PyTorch or TensorFlow.', 'NumPy is highlighted for its efficiency and usefulness in reducing the lines of code and improving the speed of the full application.', 'The purpose of the chapter is to learn how neural networks work at a very deep level, enabling a comprehensive understanding when using frameworks like PyTorch or TensorFlow.']}, {'end': 147.673, 'start': 108.443, 'title': 'Challenges in image classification', 'summary': 'Discusses the challenges faced in classifying images, from simple tasks like handwritten digits to more complex ones like mapping frames from a video game to actions, highlighting the sudden complexity and lack of direction.', 'duration': 39.23, 'highlights': ['Mapping frames from a video game to actions is suddenly complex and lacks direction.', 'Classifying images of cats and dogs is pretty simple.', 'Classifying handwritten images of handwritten digits is pretty darn simple.']}, {'end': 249.93, 'start': 147.673, 'title': 'Understanding neural network forward pass', 'summary': 'Discusses the complexity of the forward pass in a neural network, highlighting the daunting nature of its calculation and the importance of a deeper understanding for solving custom problems.', 'duration': 102.257, 'highlights': ["The forward pass of a neural network involves calculating the output information for every layer, followed by the calculation of loss to identify the network's degree of error, which can be extremely complex and daunting.", 'The calculation involves unique weights associated with input data and neurons, summed together per neuron plus a bias, and then run through an activation function for each layer.', 'The code version simplifies the process using a dot product for inputs times weights, rectified linear function for all layers, softmax activation at the end, and cross entropy loss for calculating the error.']}], 'duration': 244.505, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI5425.jpg', 'highlights': ["The forward pass of a neural network involves calculating the output information for every layer, followed by the calculation of loss to identify the network's degree of error, which can be extremely complex and daunting.", 'The chapter emphasizes creating neural networks from scratch in Python to deeply understand how neural networks work, rather than just programming a framework, with the goal of gaining a deeper understanding of frameworks like PyTorch or TensorFlow.', 'The purpose of the chapter is to learn how neural networks work at a very deep level, enabling a comprehensive understanding when using frameworks like PyTorch or TensorFlow.', 'NumPy is highlighted for its efficiency and usefulness in reducing the lines of code and improving the speed of the full application.', 'Mapping frames from a video game to actions is suddenly complex and lacks direction.', 'Classifying images of cats and dogs is pretty simple.', 'Classifying handwritten images of handwritten digits is pretty darn simple.', 'The calculation involves unique weights associated with input data and neurons, summed together per neuron plus a bias, and then run through an activation function for each layer.', 'The code version simplifies the process using a dot product for inputs times weights, rectified linear function for all layers, softmax activation at the end, and cross entropy loss for calculating the error.']}, {'end': 523.948, 'segs': [{'end': 285.45, 'src': 'embed', 'start': 249.93, 'weight': 1, 'content': [{'end': 256.591, 'text': "i encourage you to look at all of the functions and things that we're doing here to determine if any of this is really over your head,", 'start': 249.93, 'duration': 6.661}, {'end': 258.612, 'text': "because It shouldn't be right?", 'start': 256.591, 'duration': 2.021}, {'end': 260.553, 'text': "We've got very simple functions going on here.", 'start': 258.652, 'duration': 1.901}, {'end': 262.516, 'text': "We're calculating a log, sum.", 'start': 260.574, 'duration': 1.942}, {'end': 265.258, 'text': "If you don't know what log means, by the way, we're going to explain it.", 'start': 262.556, 'duration': 2.702}, {'end': 267.26, 'text': 'Sum, you should know what that means.', 'start': 265.679, 'duration': 1.581}, {'end': 271.124, 'text': "Exponential, again, if you don't know what that means, we're going to explain it, but it's very simple.", 'start': 267.58, 'duration': 3.544}, {'end': 275.408, 'text': "Dot product, again, if you don't know what that means, no worries, we'll explain it, very simple.", 'start': 271.504, 'duration': 3.904}, {'end': 280.169, 'text': 'maximum. this is just the max of whatever two values you pass here.', 'start': 275.988, 'duration': 4.181}, {'end': 282.19, 'text': 'uh, again, very simple.', 'start': 280.169, 'duration': 2.021}, {'end': 283.53, 'text': 'uh, some more dot products.', 'start': 282.19, 'duration': 1.34}, {'end': 285.45, 'text': 'a transpose, again very simple.', 'start': 283.53, 'duration': 1.92}], 'summary': 'Introduction to simple mathematical functions including log, sum, exponential, dot product, and maximum.', 'duration': 35.52, 'max_score': 249.93, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI249930.jpg'}, {'end': 394.949, 'src': 'embed', 'start': 363.371, 'weight': 4, 'content': [{'end': 366.231, 'text': "I don't expect anybody to have any background knowledge of deep learning.", 'start': 363.371, 'duration': 2.86}, {'end': 372.133, 'text': "So, if you do know things about deep learning, yes, we're gonna cover, hopefully quickly as possible, the fundamentals,", 'start': 366.291, 'duration': 5.842}, {'end': 375.334, 'text': 'just so people understand like what exactly are we aiming for here?', 'start': 372.133, 'duration': 3.201}, {'end': 381.418, 'text': 'And then the bulk of your understanding of how neural networks work is going to come from us just building these neural networks.', 'start': 376.074, 'duration': 5.344}, {'end': 383.72, 'text': "So I don't plan to spend too much time on that.", 'start': 381.839, 'duration': 1.881}, {'end': 389.485, 'text': "And if things feel a little fuzzy to you, that's probably normal, to be honest.", 'start': 383.76, 'duration': 5.725}, {'end': 394.949, 'text': "I think once you build a neural network from scratch, that is all the understanding you're going to need.", 'start': 390.185, 'duration': 4.764}], 'summary': 'Covering deep learning fundamentals, focus on building neural networks for understanding.', 'duration': 31.578, 'max_score': 363.371, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI363371.jpg'}, {'end': 458.858, 'src': 'embed', 'start': 425.206, 'weight': 0, 'content': [{'end': 429.528, 'text': 'You can use those just to kind of like spot check issues that you still find confusing.', 'start': 425.206, 'duration': 4.322}, {'end': 435.57, 'text': 'With that in mind, this series is also provided in conjunction with the Neural Networks From Scratch book.', 'start': 430.868, 'duration': 4.702}, {'end': 438.991, 'text': 'We are going to be covering the same material for the most part.', 'start': 435.63, 'duration': 3.361}, {'end': 444.633, 'text': 'The book might be a little more verbose, but the series is obviously free.', 'start': 439.091, 'duration': 5.542}, {'end': 449.054, 'text': 'The book has various prices depending on if you want the e-book, softcover, hardcover.', 'start': 445.053, 'duration': 4.001}, {'end': 450.295, 'text': 'We ship everywhere in the world.', 'start': 449.255, 'duration': 1.04}, {'end': 458.858, 'text': 'access to the book gives you access to the e-book and you know so whichever version of the book you buy, you always have access to the e-book.', 'start': 451.295, 'duration': 7.563}], 'summary': 'A series on neural networks is available for free and complements a book offered in various formats, with worldwide shipping.', 'duration': 33.652, 'max_score': 425.206, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI425206.jpg'}, {'end': 534.271, 'src': 'embed', 'start': 503.205, 'weight': 3, 'content': [{'end': 503.905, 'text': 'or vice versa.', 'start': 503.205, 'duration': 0.7}, {'end': 505.845, 'text': 'Watch the video, follow along and then read the book.', 'start': 503.925, 'duration': 1.92}, {'end': 508.546, 'text': 'Use that as kind of a way to solidify what you learned.', 'start': 505.965, 'duration': 2.581}, {'end': 513.684, 'text': "So this is a topic that you're not going to blow through this in a weekend.", 'start': 509.184, 'duration': 4.5}, {'end': 518.567, 'text': "It's going to require multiple sittings, multiple environments, and ideally multiple mediums.", 'start': 514.125, 'duration': 4.442}, {'end': 523.948, 'text': "So if you're interested in the book, you can get that at nnfs.io.", 'start': 518.767, 'duration': 5.181}, {'end': 528.049, 'text': 'So we call these neural networks because they look visually like a network.', 'start': 524.468, 'duration': 3.581}, {'end': 530.91, 'text': "You've got your neurons, which in this case are the blue circles.", 'start': 528.189, 'duration': 2.721}, {'end': 534.271, 'text': "They're connected via those orange lines.", 'start': 530.97, 'duration': 3.301}], 'summary': 'Learning about neural networks requires multiple sittings, environments, and mediums. get the book at nnfs.io.', 'duration': 31.066, 'max_score': 503.205, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI503205.jpg'}], 'start': 249.93, 'title': 'Mathematics and neural networks', 'summary': 'Covers basic mathematical functions such as log, sum, exponential, dot product, maximum, and transpose, emphasizing simplicity and accessibility. it also outlines prerequisites for viewers including programming knowledge and resources like python version and the book neural networks from scratch, offering access to the e-book, google docs draft, and early access to information.', 'chapters': [{'end': 285.45, 'start': 249.93, 'title': 'Basic mathematical functions', 'summary': 'Covers simple mathematical functions including log, sum, exponential, dot product, maximum, and transpose, emphasizing their simplicity and accessibility for understanding.', 'duration': 35.52, 'highlights': ['The chapter emphasizes the simplicity of the functions, ensuring they are accessible for understanding.', 'Functions covered include log, sum, exponential, dot product, maximum, and transpose.', 'Explanation is provided for terms such as log, exponential, and dot product to aid comprehension.']}, {'end': 523.948, 'start': 285.45, 'title': 'Neural networks prerequisites and resources', 'summary': 'Outlines the prerequisites for viewers, including a basic understanding of programming and object-oriented programming, as well as resources such as the python version and the neural networks from scratch book. the book offers access to the e-book, google docs draft, and provides an opportunity for early access to information.', 'duration': 238.498, 'highlights': ['The only expectation for viewers is to understand programming and object-oriented programming, otherwise, they will feel kind of lost.', "Viewers coming from a different programming language may find Python simple, and they can follow along in any other language as everything is low level. However, if following along in Python, it's essential to know the basics of object-oriented programming.", 'The series will cover the fundamentals of deep learning, with the bulk of understanding how neural networks work coming from building these networks, rather than spending too much time on theoretical concepts.', 'The series is provided in conjunction with the Neural Networks From Scratch book, which offers access to the e-book, Google Docs draft, and the opportunity for early access to information, making it a valuable resource for learning.', 'It is recommended to approach the topic of neural networks using multiple mediums, such as the book and the videos, to reinforce learning and understanding.']}], 'duration': 274.018, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI249930.jpg', 'highlights': ['The series is provided in conjunction with the Neural Networks From Scratch book, offering access to the e-book, Google Docs draft, and early access to information, making it a valuable resource for learning.', 'The chapter emphasizes the simplicity of the functions, ensuring they are accessible for understanding.', 'Functions covered include log, sum, exponential, dot product, maximum, and transpose.', 'It is recommended to approach the topic of neural networks using multiple mediums, such as the book and the videos, to reinforce learning and understanding.', 'The series will cover the fundamentals of deep learning, with the bulk of understanding how neural networks work coming from building these networks, rather than spending too much time on theoretical concepts.']}, {'end': 689.512, 'segs': [{'end': 609.159, 'src': 'embed', 'start': 581.957, 'weight': 0, 'content': [{'end': 587.181, 'text': 'Now the end goal of neural networks, like most machine learning, is to take some input data and produce output data that is desired.', 'start': 581.957, 'duration': 5.224}, {'end': 589.502, 'text': "In this case, we've got images of cats and dogs.", 'start': 587.321, 'duration': 2.181}, {'end': 592.704, 'text': 'We hope that we can pass it through in pixel form to our neural network.', 'start': 589.542, 'duration': 3.162}, {'end': 598.051, 'text': "And if it's a dog, then that final output neuron on top is going to be the strongest.", 'start': 593.268, 'duration': 4.783}, {'end': 601.994, 'text': "If it's a cat, then that final output neuron on the bottom will be the strongest.", 'start': 598.131, 'duration': 3.863}, {'end': 605.316, 'text': 'And we can do this by tuning the weights and biases.', 'start': 602.414, 'duration': 2.902}, {'end': 609.159, 'text': 'So all those unique weights and biases, we do that by tuning those.', 'start': 605.356, 'duration': 3.803}], 'summary': 'Neural networks process images to identify cats and dogs by tuning weights and biases.', 'duration': 27.202, 'max_score': 581.957, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI581957.jpg'}, {'end': 664.536, 'src': 'embed', 'start': 639.896, 'weight': 2, 'content': [{'end': 649.363, 'text': 'So what this ends up giving us is a huge number of uniquely tunable parameters that go into this gigantic function.', 'start': 639.896, 'duration': 9.467}, {'end': 659.51, 'text': 'So for example, with 64 by three hidden layers here, we have 9164 tunable parameters in this gigantic function.', 'start': 649.423, 'duration': 10.087}, {'end': 664.536, 'text': 'And each of those parameters impacts the outputs of the next neurons and so on.', 'start': 659.63, 'duration': 4.906}], 'summary': 'Neural network has 9164 tunable parameters for 64x3 hidden layers.', 'duration': 24.64, 'max_score': 639.896, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI639896.jpg'}], 'start': 524.468, 'title': 'Understanding neural networks', 'summary': "Explains the structure and function of neural networks, outlining their purpose in processing input data to produce desired output, such as predicting failure and training to accurately predict images of cats and dogs, while emphasizing the complexity of tuning the network's parameters.", 'chapters': [{'end': 689.512, 'start': 524.468, 'title': 'Understanding neural networks', 'summary': "Explains the structure and function of neural networks, outlining their purpose in processing input data to produce desired output, such as predicting failure and training to accurately predict images of cats and dogs, while emphasizing the complexity of tuning the network's parameters.", 'duration': 165.044, 'highlights': ['Neural networks process input data to produce desired output, such as predicting failure based on sensor data, with the ability to tune weights and biases to accurately predict data that the network has never seen before.', 'The structure of neural networks involves interconnected neurons, with a large number of uniquely tunable parameters, such as 9164 tunable parameters in a network with 64 by three hidden layers.', 'The end goal of neural networks is to take input data, like images of cats and dogs, and accurately predict the output, achieved by tuning unique weights and biases in the network.']}], 'duration': 165.044, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI524468.jpg', 'highlights': ['Neural networks process input data to produce desired output, such as predicting failure based on sensor data, with the ability to tune weights and biases to accurately predict data that the network has never seen before.', 'The end goal of neural networks is to take input data, like images of cats and dogs, and accurately predict the output, achieved by tuning unique weights and biases in the network.', 'The structure of neural networks involves interconnected neurons, with a large number of uniquely tunable parameters, such as 9164 tunable parameters in a network with 64 by three hidden layers.']}, {'end': 991.299, 'segs': [{'end': 716.595, 'src': 'embed', 'start': 689.512, 'weight': 0, 'content': [{'end': 696.236, 'text': 'since i think it would be lame to not post any code at all in this first video we are going to begin to code our first neuron.', 'start': 689.512, 'duration': 6.724}, {'end': 702.12, 'text': "but first i want to go over the version numbers real quick because, quite frankly, i'm going to forget to put it in the description.", 'start': 696.236, 'duration': 5.884}, {'end': 703.42, 'text': "so i'm using python 3.7..", 'start': 702.12, 'duration': 1.3}, {'end': 703.601, 'text': '7 numpy 1.18.', 'start': 703.42, 'duration': 0.181}, {'end': 704.061, 'text': '2 and matplotlib 3.2.', 'start': 703.601, 'duration': 0.46}, {'end': 716.595, 'text': '1. again, all of this should work very far into the future, But just in case not there are the exact versions,', 'start': 704.061, 'duration': 12.534}], 'summary': 'Coding our first neuron using python 3.7, numpy 1.18, and matplotlib 3.2.', 'duration': 27.083, 'max_score': 689.512, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI689512.jpg'}, {'end': 805.821, 'src': 'embed', 'start': 773.362, 'weight': 3, 'content': [{'end': 774.663, 'text': "It's somewhere in our neural network.", 'start': 773.362, 'duration': 1.301}, {'end': 784.751, 'text': 'Now, in this fully connected neural network, every neuron has a connection, a unique connection to every single previous neuron.', 'start': 774.863, 'duration': 9.888}, {'end': 791.356, 'text': "So let's say there's three neurons that are feeding into this neuron that we're gonna build.", 'start': 785.351, 'duration': 6.005}, {'end': 796.22, 'text': "So we don't know much about those neurons, but we know that they're outputting some values.", 'start': 791.857, 'duration': 4.363}, {'end': 800.644, 'text': "So first, their outputs become the neuron that we're coding.", 'start': 796.721, 'duration': 3.923}, {'end': 805.821, 'text': 'inputs, the neurons that were cooked anyway inputs.', 'start': 801.758, 'duration': 4.063}], 'summary': 'A fully connected neural network with 3 input neurons is discussed.', 'duration': 32.459, 'max_score': 773.362, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI773362.jpg'}, {'end': 874.318, 'src': 'embed', 'start': 846.408, 'weight': 2, 'content': [{'end': 853.05, 'text': "So you've got your inputs, your weights, and then every unique neuron has a unique bias.", 'start': 846.408, 'duration': 6.642}, {'end': 856.892, 'text': "So we're going to say bias equals three.", 'start': 853.651, 'duration': 3.241}, {'end': 866.654, 'text': 'Okay So now the first step for a neuron is to add up all the inputs times the weights plus the bias.', 'start': 857.092, 'duration': 9.562}, {'end': 869.456, 'text': 'So this is relatively simple.', 'start': 867.355, 'duration': 2.101}, {'end': 874.318, 'text': 'So in very, very raw Python, no loops required at this stage.', 'start': 869.856, 'duration': 4.462}], 'summary': 'Neurons calculate output using inputs, weights, and bias, with bias set to 3.', 'duration': 27.91, 'max_score': 846.408, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI846408.jpg'}], 'start': 689.512, 'title': 'Coding our first neuron and understanding neurons in neural networks', 'summary': 'Discusses setting up the coding environment with python 3.7, numpy 1.18.2, and matplotlib 3.2 to begin coding the first neuron and understanding the basics of coding a neuron in a densely connected feed forward, multi-layer perceptron model.', 'chapters': [{'end': 732.819, 'start': 689.512, 'title': 'Coding our first neuron', 'summary': 'Discusses setting up the coding environment with python 3.7, numpy 1.18.2, and matplotlib 3.2 to begin coding the first neuron, ensuring compatibility for future use and offering flexibility in choice of editor.', 'duration': 43.307, 'highlights': ['The chapter discusses setting up the coding environment with Python 3.7, NumPy 1.18.2, and Matplotlib 3.2 to begin coding the first neuron, ensuring compatibility for future use and offering flexibility in choice of editor.', 'The specific version numbers mentioned are Python 3.7, NumPy 1.18.2, and Matplotlib 3.2, providing exact details for compatibility and reproducibility.', 'The speaker mentions using Sublime Text as the preferred editor while acknowledging that other editors can be used, reflecting the flexibility in choice of coding environment.']}, {'end': 991.299, 'start': 732.839, 'title': 'Understanding neurons in neural networks', 'summary': 'Discusses the basics of coding a neuron in a densely connected feed forward, multi-layer perceptron model, emphasizing the concept of inputs, weights, and bias, and simplifying the understanding of neural networks from scratch.', 'duration': 258.46, 'highlights': ['The chapter simplifies the concept of neurons in a densely connected feed forward, multi-layer perceptron model, emphasizing the basics of inputs, weights, and bias, and aims to make the understanding of neural networks from scratch simple and accessible.', 'The chapter explains the concept of every neuron being connected to every single previous neuron in a fully connected neural network, highlighting the significance of unique connections and values, providing a clear understanding of the interconnectedness within neural networks.', 'The chapter elaborates on the calculation process within a neuron, emphasizing the addition of inputs multiplied by their respective weights, and the addition of bias, aiming to simplify the understanding of the computational aspect of neural networks.', 'The chapter encourages engagement and interaction by inviting questions, comments, and concerns, and offers support through Discord, fostering a collaborative learning environment for understanding neural networks from scratch.']}], 'duration': 301.787, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/Wo5dMEP_BbI/pics/Wo5dMEP_BbI689512.jpg', 'highlights': ['The chapter discusses setting up the coding environment with Python 3.7, NumPy 1.18.2, and Matplotlib 3.2 to begin coding the first neuron, ensuring compatibility for future use and offering flexibility in choice of editor.', 'The specific version numbers mentioned are Python 3.7, NumPy 1.18.2, and Matplotlib 3.2, providing exact details for compatibility and reproducibility.', 'The chapter simplifies the concept of neurons in a densely connected feed forward, multi-layer perceptron model, emphasizing the basics of inputs, weights, and bias, and aims to make the understanding of neural networks from scratch simple and accessible.', 'The chapter explains the concept of every neuron being connected to every single previous neuron in a fully connected neural network, highlighting the significance of unique connections and values, providing a clear understanding of the interconnectedness within neural networks.']}], 'highlights': ['The forward pass involves calculating output for every layer and loss, which can be complex.', 'Creating neural networks from scratch in Python emphasizes deep understanding.', 'NumPy is highlighted for efficiency and usefulness in reducing code and improving speed.', 'Neural networks process input data to produce desired output, achieved by tuning weights and biases.', 'The structure of neural networks involves interconnected neurons with a large number of tunable parameters.', 'The coding environment is set up with Python 3.7, NumPy 1.18.2, and Matplotlib 3.2 for compatibility.', 'The series covers the fundamentals of deep learning, focusing on building networks rather than theoretical concepts.', 'Approaching the topic of neural networks using multiple mediums is recommended for reinforcement.', 'The code version simplifies the process using dot product, rectified linear function, softmax activation, and cross entropy loss.', 'The purpose is to learn how neural networks work at a deep level for comprehensive understanding.']}