TensorFlow 2.0 Complete Course - Python Neural Networks for Beginners Tutorial

[@TechWithTim]

Let me know what you guys think of the course?! Took a lot of preparation and work to get this out for you guys. Hope you all enjoy and get a solid foundation in the world of machine learning :)

[@powergladius]

Eyyyyy, ur amazing

[@zyzzbodybuilding]

About 40 minutes in. Loving it dude! You have no idea how much I appreciate it. Not a fan of that haircut tho.

[@FlorinPop]

Thank you for this course Tim! I can't wait to get into it! 😃

[@shanalishams1]

Just started with the course will share my feedback once I complete this. Thank you uploading this.

[@SanataniAryavrat]

very extensive and damn good one so far...

[@phenomadit1821]

00:05 Introduction to TensorFlow 2.0 course for beginners. 02:26 Introduction to Google Collaboratory for easy machine learning setup 07:07 AI encompasses machine learning and deep learning 09:35 Neural networks use layered representation of data in machine learning. 14:12 Data is crucial in machine learning and neural networks 16:37 Features are input information and labels are output information. 21:07 Supervised learning involves guiding the model to make accurate predictions by comparing them to the actual labels 23:21 Unsupervised machine learning involves clustering data points without specific output data. 27:57 Training reinforcement models to maximize rewards in an environment. 30:00 Introduction to TensorFlow and its importance 34:36 Understanding the relation between computations and sessions in TensorFlow 36:52 Google Collaboratory allows easy access to pre-installed modules and server connection. 41:11 Importing TensorFlow in Google Collaboratory for TensorFlow 2.0 43:17 Tensors are fundamental in TensorFlow 2.0 47:58 Explanation of tensors and ranks 50:12 Understanding TensorFlow tensor shapes and ranks 54:41 Reshaping Tensors in TensorFlow 56:47 Using TF session to evaluate tensor objects 1:01:16 Different categories of machine learning algorithms 1:03:07 Linear regression for data prediction 1:07:22 Calculating the slope of a line using a triangle and dividing distances 1:09:29 Predicting values using the line of best fit 1:13:31 Overview of important Python modules like NumPy, pandas, and matplotlib 1:15:43 Predicting survival on the Titanic using TensorFlow 2.0 1:19:40 Splitting data into training and testing sets is crucial for model accuracy. 1:21:48 Separating the data for classification 1:26:09 Exploring dataset statistics and shape attributes 1:28:12 Understanding the data insights from the analysis 1:32:21 Handling categorical and numeric data in TensorFlow 1:34:39 Creating feature columns for TensorFlow model training 1:38:42 Epochs are used to feed data multiple times for better model training 1:40:55 Creating an input function for TensorFlow data set objects 1:45:19 Creating an estimator and training the model in TensorFlow 1:47:21 Explanation on how to access and interpret statistical values from a neural network model. 1:51:46 Exploring survival probabilities based on indices 1:53:52 Introduction to classification in TensorFlow 2.0 1:58:01 Data frames in TensorFlow 2.0 contain encoded species already, simplifying data preprocessing. 2:00:08 Creating input function and feature columns in TensorFlow 2.0 2:04:26 Setting up the neural network and defining the number of nodes and classes. 2:06:35 Using lambda functions to create chained functions 2:10:44 Creating a prediction function for specific flowers 2:12:46 Explaining the process of predicting on a single value 2:17:25 Clustering helps find clusters of like data points 2:19:50 Data points are assigned to clusters based on distance to centroids. 2:24:02 Understanding K means clustering 2:26:09 Hidden Markov model uses states and observations with associated probabilities. 2:30:36 Defining transition and observation probabilities in two states 2:32:56 Hidden Markov Model predicts future events based on past events 2:37:22 Explanation of transition probabilities and observation distribution in a Hidden Markov Model 2:39:31 Mismatch between TensorFlow versions 2:43:45 Hidden Markov models are used for probability-based predictions. 2:45:35 Introduction to neural networks and their working principle. 2:50:00 Designing the output layer for neural networks 2:52:19 Neural networks make predictions based on probability distributions for each class. 2:56:39 Introduction to biases as trainable parameters in neural networks 2:58:53 Neural network nodes determine values using weighted sums of connected nodes. 3:03:21 Explanation of different activation functions in neural networks 3:05:38 Sigmoid function is chosen for output neuron activation 3:10:00 Loss function measures the deviation of the neural network output from the expected output. 3:12:25 Understanding the concept of cost function and gradient descent 3:17:01 Neural networks update weights and biases to make better predictions with more data. 3:19:17 Loading and exploring the fashion amnesty dataset for training and testing neural networks. 3:23:54 Data pre processing is crucial for neural networks 3:25:54 Pre-processing images is crucial for training and testing in neural networks 3:30:26 Selecting optimizer, loss, and metrics for model compilation 3:32:33 Training and testing a neural network model in TensorFlow 2.0 3:36:51 Training with less epochs can lead to better model performance 3:39:00 Understanding predictions and probability distribution 3:43:34 TensorFlow deep learning model used for computer vision and classification tasks. 3:45:42 Images are represented by three color channels: red, green, and blue 3:50:09 Convolutional neural networks analyze features and patterns in images. 3:52:19 Convolutional neural networks use filters to identify patterns in images 3:56:49 Quantifying presence of filters using dot product 3:58:52 Understanding filter similarity in TensorFlow 2.0 4:03:09 Padding, Stride, and Pooling Operations in Convolutional Neural Networks 4:05:17 Pooling operations reduce feature map size 4:09:30 Loading and normalizing image data for neural networks 4:11:41 Understanding the input shape and layer breakdown 4:15:58 Optimizing model performance with key training strategies 4:17:59 Data augmentation is crucial for training convolutional neural networks with small datasets. 4:22:12 Utilizing pre-trained models for efficient neural network training 4:24:19 Modifying last layers of a neural network for classifying 4:28:24 Using pre-trained model, MobileNet v2, built into TensorFlow 4:30:31 Freezing the base model to prevent retraining 4:34:45 Evaluation of model with random weights before training. 4:36:58 Saving and loading models in TensorFlow 4:41:00 Natural Language Processing (NLP) is about understanding human languages through computing. 4:43:19 Sentiment analysis and text generation using natural language processing model 4:47:46 Introduction to bag of words technique in neural networks 4:49:54 Bag of words technique encodes sentences with the same representation, losing their meaning. 4:54:13 Word embeddings aim to represent similar words with similar numbers to address issues with arbitrary mappings. 4:56:25 Introduction to word embeddings in a 3D space 5:00:59 Difference between feed forward and recurrent neural networks 5:03:22 Explanation of processing words sequentially in a neural network 5:08:01 Introduction to Simple RNN and LSTM layers 5:10:29 Long Short Term Memory (LSTM) allows access to output from any previous state. 5:14:53 Padding sequences to ensure equal length for neural network input 5:17:02 Creating a neural network model for sentiment analysis 5:21:24 Evaluating model accuracy and preparing for predictions 5:23:49 Explanation of padding and sequence processing in TensorFlow 2.0 5:28:20 Analyzing sentiment impact on prediction accuracy 5:30:27 Training neural network to generate text sequences 5:34:48 Creating mapping from characters to indices 5:37:09 Creating training examples for TensorFlow neural network model 5:41:53 Batching and model building process in TensorFlow 2.0 5:44:07 Setting model parameters and layers in TensorFlow 2.0 5:49:05 Explaining model predictions for each element in batch and sequence length 5:51:26 The model outputs a tensor for each training example, and we need to create our own loss function to determine its performance. 5:56:05 Training neural networks with varying epochs for performance evaluation 5:58:29 Generating output sequences using TensorFlow model 6:02:53 Processing steps for text data in TensorFlow 2.0 6:05:05 Building and training the model with different batch sizes and checkpoints 6:09:25 Reinforcement learning involves an agent exploring an environment to achieve objectives. 6:11:43 States, Actions, and Rewards in Reinforcement Learning 6:16:24 Q matrix represents predicted rewards for actions in states. 6:18:43 Maximize agent's reward in the environment 6:23:21 Introducing exploration in reinforcement learning 6:25:26 Balancing Q table and random actions in Q learning algorithm 6:30:03 Discount factor helps in factoring future rewards into the equation for finding the best action in the next state. 6:32:16 Introduction to OpenAI Gym for training reinforcement learning models 6:36:46 Introduction to navigating a frozen lake environment using q learning. 6:38:54 Max steps and learning rate in reinforcement learning 6:43:05 Training the agent using Q-learning algorithm 6:45:18 Training process involves adjusting epsilon and monitoring reward progress. 6:49:39 Focus on a specific area in machine learning or AI for deeper learning. 6:51:47 Largest open source machine learning course in the world focused on TensorFlow and Python. Crafted by Merlin AI.

[@DesignDazzler5]

ur a savior brother , thanks

[@musicsimp536]

Absolute legend 🫂

[@thechairisfloatingaround]

thanks bro ur a lifesaver

[@net.5503]

⌨️ Module 1: Machine Learning Fundamentals (00:03:25) ⌨️ Module 2: Introduction to TensorFlow (00:30:08) ⌨️ Module 3: Core Learning Algorithms (01:00:00) ⌨️ Module 4: Neural Networks with TensorFlow (02:45:39) ⌨️ Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10) ⌨️ Module 6: Natural Language Processing with RNNs (04:40:44) ⌨️ Module 7: Reinforcement Learning with Q-Learning (06:08:00) ⌨️ Module 8: Conclusion and Next Steps (06:48:24)

[@harsht4345]

just copied this from the description lol

[@AadityaMankarXv7]

@@harsht4345 For phone users mate.

[@CodeZeroSix]

Thnxx man

[@오재화-d6l]

Thanks

[@sandeshadhikari2889]

Can i learn machine learning without having a laptop with dedicated Graphics card?? Please help

[@aaronpaul2550]

I think this course gives a chance to anyone who wants to learn machine learning in a fast and free way. And save a bunch of time looking at papers and library literature. This course is gradual. There is a clear understanding of everything from linear regression to reinforcement learning, and even the example programs are fully described and annotated. The people who made and designed this course are very thoughtful and selfless sharing and deserve huge applause. Thank you very much.

[@bohaning]

Hey, I'd like to introduce you to my AI learning tool, Coursnap, designed for youtube courses! It provides course outlines and shorts, allowing you to grasp the essence of 1-hour in just 5 minutes. Give it a try and supercharge your learning efficiency!

[@SomeshB]

CAUTION: The Vido is outdated, you can use the video for concepts but code wise TensorFlow has deprecated many modules that are used in the code he mentioned.

[@gdfra4733]

i'm at 2 hours and the only thing i had a problem with was tensorflow.compat which now is tensorflow._api.v2.compat.v2

[@vjndr32]

@@gdfra4733 you might be using older version of tensorflow. i'm on 2.16.1 on my local machine and not able run even the linear regression

[@EJYIEI]

@@vjndr32 in my case i decided to use keras models instead of estimators since the official tensorflow page itself has a tutorial to migrate

[@KhalidKassim-gc6mj]

That’s y he said to use the same 2.0 version 🤦‍♂️

[@LandonCummings1]

@@EJYIEI hey im at that point where I want to use keras models instead of estimators but I cant figure out the tutorial for a linearregression model any chance you could share your code for that section?

[@thesral96]

Please consider adding Chapters to the KZbin Progress Bar so that the information is easier to find later on.

[@nickfiction5507]

⌨️ Module 1: Machine Learning Fundamentals (00:03:25) ⌨️ Module 2: Introduction to TensorFlow (00:30:08) ⌨️ Module 3: Core Learning Algorithms (01:00:00) ⌨️ Module 4: Neural Networks with TensorFlow (02:45:39) ⌨️ Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10) ⌨️ Module 6: Natural Language Processing with RNNs (04:40:44) ⌨️ Module 7: Reinforcement Learning with Q-Learning (06:08:00) ⌨️ Module 8: Conclusion and Next Steps (06:48:24)

[@acosmic7841]

Thanks

[@aridorjoskowich7283]

​@@nickfiction5507 doing God's work

[@NikhilYadav-ji8rm]

Timestamps for all the different core learning algorithms, Linear Regression (01:00:00) Classification (01:54:00) K-Means Clustering (02:17:07) Hidden Markov Models (02:24:56)

[@emberleona6671]

@User Account Karen!

[@filipo4114]

03:43:10 - Convolutional Neural Networks

[@goksuceylan8844]

@User Account ur mom

[@prasaddalavi9683]

Hey just to confirm, are you sure the 1:00:00 is the linear regression and not linear classification. i am not able to get this. we are classifying whether it will be survived or not. based on the input data. can some one please help with this

[@leonardodalcegio4763]

@@prasaddalavi9683 I am also in doubt about that

[@UthacalthingTymbrimi]

1:19:30 In case anyone is curious, in the Titanic Dataset, "parch" is "Parents/Children" (ie: was this person travelling with other family members), and "fare" is the price paid for their ticket (which may include travelling costs for other people they were travelling with, family or not).

[@SandSeppel]

thank you so much

[@user-ze7sj4qy6q]

thank u i figured what fare was but i had no idea abt parch n it was bothering me lol

[@ramazad1363]

48:00 rank 50:00 shape 52:00 change in shape 55:10 types of tensors 56:30 evaluating Tensors 57:25 sources 57:40 practice 1:00:00 tensorflow core learning algorithms 1:02:40 linear regression 1:13:00 setup and import 1:15:40 data

[@techstuff7568]

'I'm sorry I'm talking a lot but...' Bro, it's a 7 hour TensorFlow tutorial, I didn't expect anything less! Awesome tutorial, thanks man

[@itjustmemyselfandi]

Can I ask how long it took to learn and watch this video?

[@saicharansigiri2964]

@@itjustmemyselfandi 5days

[@sorvex9]

@@itjustmemyselfandi 1 day

[@aparupganguly01]

@@itjustmemyselfandi 9 months

[@pe....]

@@itjustmemyselfandi 4 hours

[@henryly213]

Thing I needed to update as I went thru the course (will update as I go) 1:13:12 - Package change to scikit-learn : !pip install -q scikit-learn

[@dawidkoperek7917]

Thank you very much. The goat

[@someatuffs]

⌨ (00:03:25) Module 1: Machine Learning Fundamentals ⌨ (00:30:08) Module 2: Introduction to TensorFlow ⌨ (01:00:00) Module 3: Core Learning Algorithms ⌨ (02:45:39) Module 4: Neural Networks with TensorFlow ⌨ (03:43:10) Module 5: Deep Computer Vision - Convolutional Neural Networks ⌨ (04:40:44) Module 6: Natural Language Processing with RNNs ⌨ (06:08:00) Module 7: Reinforcement Learning with Q-Learning ⌨ (06:48:24) Module 8: Conclusion and Next Steps Progress: 04:00

[@toihirhalim]

this made me understand what I've been learning for 2 semesters

[@andyh964]

This video is gold. I am a MSc student in AI and I literally use this video as a reference to understand some topics that are poorly explained in the modules. I've watched 5/7 hours

[@damascenoalisson]

Just a comment, on 3:36:44, when you train the network again it's not re-training from scratch but instead using the weights it already had. Unless you manually reset the graph, you'll be training for the sum of all epochs you used the fit function (like 10 + 8 + 1 epochs) To avoid this problem you should use something like keras.backend.clear_session() or tf.reset_default_graph() between tests with hyperparameters 😉

[@cwlrs4944]

Mm thought that was the case. Wasn't starting from the ~80% accuracy from the first epoch of the latter training runs.

[@lawrencegranda7759]

Another way is just to rebuild the model.

[@ufukdemiray6176]

this was painful to watch yeah.. i know he's doing his best to show stuff but he's pretty much a beginner too

[@ScriptureFirst]

I LOVE this firehose format of SPRINT-crawl-walk. Everyone thinks they need to crawl-walk-run & that’s crap. I like your style dude.

[@leixun]

*My takeaways:* 1. TensorFlow has two main components: graph and session 33:05 2. We can rebuild a model and change its original parameters 5:56:31 3. Reinforcement learning with Q-Learning 6:08:00

[@beastkidoooo]

very good course tim . I am 12 and have finished ur course upto neural networks ur a great teacher , ignore the bad comments cause the people who post these comments are failures that are jealous of u , so never ever give up !!

[@porterneon]

parch: The dataset defines family relations in this way... Parent = mother, father Child = daughter, son, stepdaughter, stepson Some children travelled only with a nanny, therefore parch=0 for them.

[@robertboles7418]

Par/Ch : Parents / Children (aboard)

[@nadiakacem24]

⭐️ Course Contents ⭐️ ⌨️ Module 1: Machine Learning Fundamentals (00:03:25) ⌨️ Module 2: Introduction to TensorFlow (00:30:08) ⌨️ Module 3: Core Learning Algorithms (01:00:00) ⌨️ Module 4: Neural Networks with TensorFlow (02:45:39) ⌨️ Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10) ⌨️ Module 6: Natural Language Processing with RNNs (04:40:44) ⌨️ Module 7: Reinforcement Learning with Q-Learning (06:08:00) ⌨️ Module 8: Conclusion and Next Steps (06:48:24)

[@DrRussell]

Just started. Know nothing so can’t contribute yet but wanted to thank you for advancing humanity. You may have just given my life a purpose

[@benlaurent3102]

How’s it been going? Are you still doing machine learning?

[@cutyoursoul4398]

life has no purpose

[@thesickbeat]

@@cutyoursoul4398 Said the atheist.

[@cutyoursoul4398]

@@thesickbeat not atheist, that's just the Truth

[@thesickbeat]

@@cutyoursoul4398 Its your truth. Not the truth.

[@manikandans2030]

Run time 3:37:00 - I think we have to compile the model every time before we do a fit. Otherwise it just memorize the previous epochs and use it for next iterations. In this case I believe that 92% accuracy of 1 epochs is the same as the addition of previous epochs i.e 10+8+1 = 19 epochs

[@WalkerSuper900]

I agree 100%. He was overfitting the model even more.

[@ScriptureFirst]

I typically hate narrative talk alongside & prefer scripted tutorials, but you’ve spoken very clearly & concisely while extemporaneously. Very well done! 🙏🏼

[@thesultan1212]

This video is pure gold, the guy explains really well. Learned more from this than payed courses. Thanks so much, keep it up!

[@CivilSurveyor]

Hello Sir.. I need a program, vba, excel sheet, or anything else.. In which I do add may data in Numaric form that is 1,2,3 etc and then that data plot or draw in AutoCAD with one click

[@Trixz-the]

@Dario Argies relax pal

[@yungrabobank4691]

For people wanting to understand the basic idea behind Neural Networks, 3BlueOneBrown's video is a nice addition to your introduction! It helped me understand the topics and coding Tim discussed a lot better

[@danielleivy8180]

Also Stanford has their full CS229 course online as well - along with lecture notes. :)

[@pabloa.2586]

@@danielleivy8180 where can i find that course? thanks in advance

[@danielleivy8180]

@@pabloa.2586 kzbin.info/www/bejne/l5rVlHSoqtuhgc0si=02lMoL958AjkXkyy

[@redwings5576]

I've been going back and forth between python to game development... but I haven't actually learned anything and now I'm here for machine learning, and before this I started hacking course. Everything unfinished, and just the thought of what I would be able to do when I get good at either of these is what makes me want to learn them, but I can't really stick to one.

[@abhinavyadav789]

Happens with everyone . I went with flask to build a website , then api , left it , did a but of numpy , now here . Im just looking to find something that interests me enough to make me stick to it for a longer time, you should keep looking for something that might interest you so that you stick with it longer !

[@redwings5576]

@@abhinavyadav789 The thing I really want to do, is what I don't have support with of any kind. So, it's like, I'm just trying to find something, in places where I know I won't find it.

[@AllenJose-p3r]

🎯 Key Takeaways for quick navigation: 00:00 *🎓 Introduction to Course and Audience* - Aimed at beginners in machine learning and artificial intelligence with basic programming knowledge. 03:16 *📚 Course Structure and Resources* - Course breakdown, starting with machine learning and AI basics. 10:16 *🤖 Understanding Artificial Intelligence, Machine Learning, and Neural Networks* - Definition of Artificial Intelligence (AI) as automating intellectual tasks. 14:25 *📊 Importance of Data in Machine Learning* - Example dataset creation for student grades. 16:17 *📊 Features and Labels Basics* - Features are input information for machine learning models. 17:42 *📈 Importance of Data in Machine Learning* - Data is essential for creating machine learning models. 19:35 *🧠 Types of Machine Learning: Supervised Learning* - Supervised learning involves having both features and labels. 22:43 *🌐 Types of Machine Learning: Unsupervised Learning* - Unsupervised learning deals with only features and no labels. 25:25 *🤖 Types of Machine Learning: Reinforcement Learning* - Reinforcement learning involves an agent, environment, and reward. 30:10 *🧠 Introduction to TensorFlow and Module Structure* - TensorFlow is an open-source machine learning library by Google. 31:38 *🚀 What can be done with TensorFlow?* - TensorFlow supports various machine learning tasks and neural networks. 32:54 *🧠 TensorFlow Overview* - TensorFlow provides a library of tools for machine learning applications. 36:02 *🚀 Getting Started with Google Collaboratory* - Google Collaboratory allows using Jupyter Notebooks in the cloud. - Specify TensorFlow version in Collaboratory with `%tensorflow_version 2.x`. 43:29 *🧮 Understanding Tensors* - Tensors are a generalization of vectors and matrices to potentially higher dimensions. 47:42 *📏 Rank and Degree of Tensors* 49:03 *📊 Understanding Tensor Rank and Shape* 52:13 *🔄 Reshaping Tensors in TensorFlow* 55:30 *🧠 Types of Tensors in TensorFlow* 56:51 *🔄 Evaluating Tensors using Sessions* 01:00:06 *🤖 Introduction to Core Machine Learning Algorithms* 01:04:20 *📈 Linear Regression Basics* 01:05:40 *🔄 Using Linear Regression in Prediction* 01:10:33 *📊 Linear Regression in Three Dimensions* 01:12:29 *🔍 Examples of Linear Regression* 01:15:48 *🛳️ Titanic Dataset for Linear Regression* 01:19:06 *📊 Data Preparation: Understanding Columns in Dataset* 01:20:02 *📊 Data Preparation: Creating Training and Testing Sets* 01:21:19 *📊 Data Exploration: Pandas DataFrames and Descriptive Stats* 01:26:57 *📊 Data Visualization: Creating Histograms and Plots* 01:29:47 *📊 Data Understanding: Training and Testing Sets Analysis* 01:30:14 *📊 Feature Columns: Categorical and Numeric Data* 01:33:55 *🧮 Feature Columns for Categorical Data* 01:36:14 *📊 Feature Columns for Numeric Data* 01:37:08 *🔄 Training Process Overview* 01:40:20 *🤖 Input Function Creation* 01:45:02 *🤖 Creating Linear Estimator* 01:46:25 *🚂 Model Training Process* 01:47:48 *📈 Model Evaluation and Predictions* 01:55:04 *📊 Introduction to Classification* 01:57:25 *🗃️ Loading and Preparing Dataset* 02:00:08 *🔄 Input Function* 02:01:36 *🧮 Feature Columns* 02:03:33 *🧠 Building a Deep Neural Network Classifier* 02:04:26 *🧠 Neural Network Architecture* 02:04:56 *🤖 Training the Model* 02:07:43 *🧾 Training Output and Evaluation* 02:09:58 *🔄 Model Evaluation and Prediction* 02:11:53 *📊 Predictions on New Data* 02:17:20 *🤔 Introduction to Clustering (K-Means)* 02:20:07 *🌐 K-Means Clustering Overview* 02:25:14 *📊 Hidden Markov Models Introduction* 02:28:22 *🎲 States, Observations, and Transitions* 02:33:56 *🔍 Purpose of Hidden Markov Models* 02:35:44 *🌡️ Hidden Markov Model Introduction,* 02:37:09 *📊 TensorFlow Probability Distributions,* 02:38:33 *📉 Building the Hidden Markov Model,* 02:41:40 *🔄 Modifying Probabilities and Observing Changes,* 02:45:47 *🧠 Introduction to Neural Networks,* 02:50:45 *🧠 Neural Network Layers and Output Design* - Single output neuron with a value between 0 and 1 for binary classification. - Multiple output neurons for predicting probabilities in a classification task. 02:53:07 *🔗 Hidden Layer in Neural Networks* 02:54:28 *🌐 Connectivity: Weights and Biases in Neural Networks* 03:00:25 *⚙️ Weighted Sum, Bias, and Information Flow* 03:02:39 *🔄 Activation Functions in Neural Networks* 03:06:50 *URL](kzbin.info/www/bejne/qoG8m2ace696oM0) 🧠 Activation Functions* 03:08:11 *URL](kzbin.info/www/bejne/qoG8m2ace696oM0) 📊 Moving to Higher Dimensions* 03:09:32 *URL](kzbin.info/www/bejne/qoG8m2ace696oM0) 📉 Loss Function Basics* 03:12:50 *URL](kzbin.info/www/bejne/qoG8m2ace696oM0) ⚙️ Optimizing with Gradient Descent* 03:18:01 *URL](kzbin.info/www/bejne/qoG8m2ace696oM0) 🔄 Building the First Neural Network* 03:23:08 *🖼️ Image and Label Exploration* 03:24:27 *🔄 Data Pre-processing* 03:27:13 *🧠 Model Creation* 03:30:04 *🤖 Compiling the Model* 03:32:20 *⚙️ Training the Model* 03:35:05 *🧪 Testing and Evaluating the Model* 03:38:19 *🖼️ Overview of Image Prediction* 03:39:16 *🧠 Understanding Predictions with Arrays* 03:40:07 *🥿 Decoding Predictions to Class Names* 03:41:25 *🤖 Verifying Predictions Script* 03:43:17 *🌐 Introduction to Convolutional Neural Networks (CNN)* 03:45:11 *🖼️ Understanding Image Data Dimensions* 03:47:01 *🔄 Global vs. Local Patterns in Neural Networks* 03:51:11 *📊 Convolutional Layer Output Feature Maps* 03:54:23 *🎨 Convolutional Neural Network Overview* 03:55:19 *🖼️ Looking for Filters in Images* 03:56:40 *📊 Dot Product and Feature Maps* 04:00:23 *🔄 Padding, Stride, and Computational Efficiency* 04:04:59 *🏞️ Pooling Operations* 04:08:10 *🚀 Building a Convolutional Neural Network with Keras* 04:09:30 *🖼️ Loading CIFAR-10 Dataset and Normalization* 04:12:40 *🧱 Convolutional Base Summary* 04:14:51 *🧠 Adding Dense Layers for Classification* 04:15:49 *🎓 Model Training and Evaluation* 04:18:37 *🔄 Data Augmentation* 04:22:42 *🤖 Using Pre-trained Models* 04:24:10 *🤖 Using Pre-trained Models* 04:25:33 *📊 Loading and Preprocessing Data* 04:26:56 *🖼️ Image Reshaping and Scaling* 04:29:43 *🧠 Picking a Pre-trained Model* 04:31:55 *🧊 Freezing the Base Model* 04:33:19 *🏗️ Adding Custom Classifier* 04:34:39 *🎓 Model Compilation and Evaluation* 04:36:58 *🚂 Training the Model* 04:37:57 *💾 Saving and Loading Models* 04:38:24 *🔍 Introduction to Object Detection* 04:38:50 *🧠 Understanding TensorFlow and Facial Recognition* 04:41:09 *🗣️ Natural Language Processing with Recurrent Neural Networks* 04:42:35 *📈 Applications of Recurrent Neural Networks* 04:44:24 *📊 Challenges in Textual Data Processing* 04:51:55 *🔠 Issues with Direct Word-to-Integer Encoding* 04:54:39 *📊 Understanding Word Representation Challenges* 04:55:32 *🛠️ Word Embeddings Overview and Visualization* 04:58:44 *🧠 Word Embeddings as a Layer in Neural Networks* 04:59:41 *🔢 Preparing Textual Data for Neural Networks* 05:02:26 *🔄 Unraveling Recurrent Neural Network Layers* 05:09:49 *🚀 Long Short-Term Memory (LSTM) Layers* 05:11:12 *🧠 Understanding Long Short-Term Memory (LSTM)* 05:12:34 *🎥 Sentiment Analysis on Movie Reviews* 05:13:55 *📊 Data Preprocessing for Neural Network Input* 05:17:11 *🧠 Building and Compiling the LSTM Model* 05:21:43 *📈 Model Evaluation and Results* 05:24:19 *🧐 Making Predictions with the Trained Model* 05:26:44 *🤖 Overview of Text Processing* 05:28:09 *📈 Sentiment Analysis Example* 05:30:23 *🎭 Recurrent Neural Network for Text Generation* 05:31:48 *🧠 Data Loading and Preprocessing* 05:34:06 *🧮 Encoding Characters and Creating Functions* 05:37:18 *🚀 Creating Training Examples* 05:38:39 *🔄 Mapping Sequences and Creating Batches* 05:41:23 *🏗️ Building the Model* 05:42:46 *🏗️ Building the Model Architecture* 05:47:13 *📉 Creating a Loss Function* 05:55:57 *🚂 Compiling and Training the Model* 05:57:23 *🔄 Rebuilding Model for Inference* 05:59:41 *🧠 Understanding Text Generation with RNNs* 06:07:20 *🤖 Recap and Guidance on Complex ML Concepts* 06:08:18 *🎮 Introduction to Reinforcement Learning* 06:09:42 *🔄 Key Concepts: Environment, Agent, State, Action, Reward* 06:15:10 *🧠 Introduction to Q-Learning in Reinforcement Learning* 06:15:36 *🤖 Q-Learning Introduction* 06:17:50 *🎨 Q-Learning Example on Whiteboard* 06:19:14 *🕹️ Navigating the Environment and Learning the Q-Table* 06:25:15 *🔄 Learning the Q-Table - Constants and Update Formula* 06:31:40 *🧠 Q-Learning and OpenAI Gym Introduction* 06:32:35 *🎮 OpenAI Gym Environment Setup* 06:33:58 *📊 Constants and Environment Setup* 06:35:21 *🔄 Picking Actions in Q-Learning* 06:42:25 *🔄 Updating Q-Values* 06:44:39 *🚀 Training Q-Learning Model* 06:45:58 *📈 Training Results and Graph* 06:47:17 *🤖 Q-Learning Example Conclusion* 06:48:13 *🏁 Conclusion of Reinforcement Learning Module* 06:48:41 *🚀 Next Steps and Further Learning Recommendations* 06:50:26 *🎓 Advice for Specialization and General Exploration* 06:51:47 *🏆 Course Conclusion and Call to Action* Made with HARPA AI

[@pallavijog912]

At 3:37:00, when you said with less epochs, you are getting better test results. Which is actually not the case. You first run for 10 epochs, your weights got updated. Then again you run 8 epochs, your weights improved from previous values onwards.. so that eventually makes 18 epochs.. then you run for 1 epoch, which makes it 19 epochs.. so in this case, after 19th epoch, your accuracy on test data is increased.

[@yashvander-bamel]

I was about to write the same thing...seems like I'm not the only one who noticed :)

[@Adi-zs7bm]

Module 1: Machine Learning Fundamentals (00:03:25) Module 2: Introduction to TensorFlow (00:30:08) Module 3: Core Learning Algorithms (01:00:00) Module 4: Neural Networks with TensorFlow (02:45:39) Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10) Module 6: Natural Language Processing with RNNs (04:40:44) Module 7: Reinforcement Learning with Q-Learning (06:08:00) Module 8: Conclusion and Next Steps (06:48:24)

[@skviknesh]

1:02:44 "Do not Memorize just Understand" - made my mind to stay "calm". Felt to thank at that time frame... "Thank You!"

[@Pinocciochannel]

Well im not the best at python.. But its my favorite out of all the language i know. So whenever i dont understand something i be like chill its just python. That helps at least to some extend.

[@GabrielAzevedo_11]

I thought the same, it gave me a relief.

[@dlerner97]

Okay I'm not completely sure about this so take it with a grain of salt but I don't think you're hyperparameter/epoch tuning at 3:37:00 is doing what you expect. With jupyter notebooks, it saves the models and each time you run an epoch, it continues tuning the previous weights. In order to really display epoch differences, you need to restart the runtime and repeat the process. If you notice, each time you run the code, the "first epoch accuracy" increases significantly. The first time you ran it, the accuracy was 83% after the first epoch. After the 10th, it was 90.6%. Then, for the next iteration (8 epochs), the accuracy was 91.2% after the first epoch. Then, when running on just a single epoch, it started at 93%. Likely this is because the model continued to train an additional 9 epochs. So, in fact, the single epoch data is ironically quite overfitting.

[@devloper_hs]

FOR TENSORFLOW 2.0 For running seesions at : 57:03 with tf.compat.v1.Session() as sess: print(tensor0.eval())

[@yaswanthravuri8450]

Classical programming : Answers=f(Data, Rules) Machine learning : Rules = f(Data, Answers) That cleared all my questions on difference between classical programming and machine learning . Thanks for that 😇😇

[@rishabhgarg1445]

Just wanted to add one small detail that in Module 4, while training the model on 1 epoch after training it for sometime, what actually happens in ipython notebooks is that they continue training on above of the previously trained model. So, that is why we got pretty high accuracy for one epoch, but technically that accuracy we got was not just from one epoch.

[@eric9964]

Are you sure? Adding on one epoch to that model made a significant jump from its previous accuracy. I don't believe this is the case.

[@lawrencegranda7759]

I agree. He did not restart/rebuild the model, so it just kept training using the previous weights

[@RodrigoLobatorodrigo]

@@lawrencegranda7759 I was watching this and even though I am totally newbie I also noticed that the training was simply continuing, not starting from scratch.

[@humanbeing2282]

1:19:41 someone else has to have made a comment about this but fare refers to the price you paid for a ticket. As in the amount they paid to board the ship for the voyage. It’s a general term that broadly speaking means amount that you contributed economically to partake in or embark on an activity. You could feasibly replace fare with “price of ticket” or “entrance fee” and it would mean the same thing. Fare has a slightly different connotation but in any practical way it’s a synonym for cost of a thing. It’s notably not a tensor flow specific term.

[@gadi800]

Despite being lost in the RNN part haha, this tutorial was great! I really appreciate your hard work and you've done great in simplifying explanations. Well done! It's programmers like yourself that make it possible for anybody to learn programming and that is a great thing. I hope to see more courses from you in the future!

[@networkserpent5155]

I only watched until 20 minutes today, but I can say this really helped me get a theortical grasp on machine learning in general. I just though machine learning is just predicting data but today i learned that in order for it to do that it uses an algorithm that make rules than follows them and gives back the label (output) thank you sm!!

[@jdcrunchman999]

You should NOT expect us to "look up" some of these parameters, instead you should explain them, if not, then please give reference of where we should look them up. I'm referring to the video position that ended around 3.31. once you started explaining the math, I started to understand. but I know your viewers are not 2nd year calculus. but so far, this is the best video out there that explains Tensor flow.

[@marufm8195]

Just finished the tutorial, it's really well made and an amazing intro to ML concepts. I'm really excited to explore this further thankyou so much Tim.

[@bohaning]

Hey, I'd like to introduce you to my AI learning tool, Coursnap, designed for youtube courses! It provides course outlines and shorts, allowing you to grasp the essence of 1-hour in just 5 minutes. Give it a try and supercharge your learning efficiency!

[@johnsonamodu77]

Parch represents number of parents and/or children passengers onboard the ship with (1:17:49) Fare represents the fare price. The ticket price (1:18:05)

[@DhruvPatel11]

Thanks, man, I was having difficulties learning core concepts of ML for a long time but this video cleared all my queries and now I understand everything which you've explained. Thanks again for making this video. It helped a lot

[@yousefwaelsalehelsaidkhalil]

Wow, Amazing. Eventually I lost the hole to continue in machine learning but you just had gave me a road to run on for a long amount of time!!

[@bobmimiaga]

Tim, I know you've heard this before, but this was a very well done course on Tensorflow basics and Machine Learning. This is my first online course I've taken and glad it was yours! Thanks for the time you put into this course which will help countless programmers and adventurers interested in this fascinating field.

[@mariuspopovici4296]

Fare would be the amount they paid for the trip / ticket price. Parch is # of Parents/Children aboard.

[@vovin8132]

Yeah I was thinking that fare was a function of cabin class (base value) and destination (length on board).

[@waiitwhaat]

3:36:25 The accuracy keeps going up is because everytime you execute the codeblock, it takes the trained model and runs it for the specified number of epochs, since .fit() is being called on the same variable 'model'. To start training it with a fresh model, just re-initialize the variable 'model' by executing the code block with 'model = keras.Sequential([...])'.

[@acidnynex]

Good work, I appreciate you trying to teach the masses. However, the first example is not linear regression, it is binomial logistic regression and doesn't really represent what you explain earlier in the first part of the video. Perhaps the housing price data set or another data set would be a good example for this, with binomial logistic regression as a second step that then leads into multinomial logistic regression.

[@saikatraj3113]

TIME STAMP: ⌨️ Module 1: Machine Learning Fundamentals (00:03:25​) ⌨️ Module 2: Introduction to TensorFlow (00:30:08​) ⌨️ Module 3: Core Learning Algorithms (01:00:00​) ⌨️ Module 4: Neural Networks with TensorFlow (02:45:39​) ⌨️ Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10​) ⌨️ Module 6: Natural Language Processing with RNNs (04:40:44​) ⌨️ Module 7: Reinforcement Learning with Q-Learning (06:08:00​) ⌨️ Module 8: Conclusion and Next Steps (06:48:24​)

[@rameshbalakrishnan_meoooow]

ok when he said at the start im going to explain everything I didn't believe. he literally explained everything used. like even summation, SD etc., real pro. damn I'm never gonna get good like this guy.

[@GeekTutorials1]

Mate, this was very cool. I'd never heard of it before, but coming from a Python background, I found this very helpful. Keep up the great work! Looking forward to what else you have on offer.

[@garzj]

The only thing that bothers me is the way that he draws pacman...

[@raspberrypi4970]

Try OceanSDK/Leap2 from D-Wave

[@ninaddesai5105]

This is aweome. I paid for a course online - but could not understand lots of things .. just was able to clear all confusing concepts in this video. Thanks mate

[@mohdabdulrahman4210]

it's only 30 minutes and I'm already loving it

[@ScriptureFirst]

Thank you for putting comments in each line. Many people skip this level of detail. I love that you’ve wrapped this in comments. 🙏🏼

[@dannloloy]

He may not be the greatest teacher tbh (but he is included to those who are great) but his commitment to teaching is undeniable! Thank you sir.

[@michaelmarinos]

Great Video (even if i havent finished it yet)! The example in 1:00:10 is NOT regression is Classification. You have 2 classes (survived or not) and you try to classify the passengers. In other words, the result would always be a probability, you cannot use the same methodology to predict the age, for example .

[@ci9vt]

The second argument to tf.Variable() is trainable not dtype, so when you set string = tf.Variable('some string', tf.string), you set string.trainable to tf.string. You can verify it by printing string.trainable.

[@mom4839]

Where is the subtitle??

[@masudulalam2515]

what is string.trainable?what is the purpose of it?I'm real noob here,help me out!!

[@sangramjitchakraborty7845]

@@masudulalam2515 it sets the variable as trainable or not. Trainable variables are updated during training. Like weights and biases.

[@sandeshadhikari2889]

Can i learn machine learning without having a laptop with dedicated Graphics card?? Please help( i am going to buy a laptop with low budget)

[@fernandogamdev]

I just don't know how to thank you ALL YOUR EFFORT on doing that video! All the content and the explanation! It's just mind blowing! I am eternally grateful!

[@jedi4ever]

I really, really enjoyed this tutorial . It takes the time to explain soo many aspects and has a great build up. Well done!

[@jamesmuthama1750]

If you're a complete beginner, ChatGPT explains the difficut concepts so well

[@owusukwakumoses99]

It's 2022 and this video is as relevant as ever. I didn't really follow up in the latter parts but you really did a good job explaining stuff. Thanks!!!

[@freecodecamp]

Click the "JOIN" button below the video to support freeCodeCamp.org!

[@soyoutube22]

In module 3 at 2:36:20, it's accidentally backwards in the ipynb file. "initial_distribution = tfd.Categorical(probs=[0.2, 0.8])" should be [0.8, 0.2] like in the video. Actually, all the numbers in that code block are mismatched from the video and don't match the weather diagram.

[@soyoutube22]

Also "!pip install tensorflow_probability==0.8.0rc0 --user --upgrade" will no longer work, it needs to be "!pip install tensorflow_probability --user --upgrade" and then doing a FACTORY RESET RUNTIME and clicking the reconnect button.

[@xewwwqxxa5081]

SUBTITLES on this video would have been great

[@gibsosmart]

Thanks to NICK FICTION who formulated it ⌨️ Module 1: Machine Learning Fundamentals (00:03:25) ⌨️ Module 2: Introduction to TensorFlow (00:30:08) ⌨️ Module 3: Core Learning Algorithms (01:00:00) ⌨️ Module 4: Neural Networks with TensorFlow (02:45:39) ⌨️ Module 5: Deep Computer Vision - Convolutional Neural Networks (03:43:10) ⌨️ Module 6: Natural Language Processing with RNNs (04:40:44) ⌨️ Module 7: Reinforcement Learning with Q-Learning (06:08:00) ⌨️ Module 8: Conclusion and Next Steps (06:48:24)

[@puspamadak]

This video is a must-watch for beginners getting into machine learning. I wish I had seen this video before. I have never got a better understanding of these topics and the differences between the terms AI, Machine Learning, etc. Thank you, sir, for your efforts. I am in class 12, and there is Linear Regression in Mathematics, but I haven't even thought that it can be used in ML also.

[@shdnas6695]

Just curious to know, what are u doing now dude? i mean in programming area

[@Luxcium]

46:54 A tensor of degree 0 is a scalar, which is a single numerical value. Therefore, the shape of a tensor of degree 0 is an empty tuple, denoted by '()'.

[@vanishingentropy6488]

Loved it! Great tutorial covering a lot of areas. TechWithTim's explanation and the epic examples help open up the field to beginners like me, and the 7 hours were super-interesting!

[@rohanmanchanda5250]

Well, I'm Batman.

[@lucasbishop1945]

​@@rohanmanchanda5250 Well, I'm Spiderman.

[@andrewwheeler8099]

I'm only 35 minutes in but just have to make a comment and let you know I love the way you present this and your communication is great. I appreciate your time making this and I'm very glad you did! Thanks Dawg

[@gioannguyen4213]

I've been in this course for over 2 hours. I think this is a good point to start with as it can equip beginners with a big picture, together with quick (but enough) explanations on the terms such as ML, layer, etc. Although some of the codes provided in the GG collab couldn't run properly in 2024 (right now), I suggest watching the video to have a grasp on what might happen and practice them later (maybe in another course, or if you can figure out somehow to execute these codes). Happy learning!

[@BernardLawes]

One of the best executed courses on KZbin. Very well done!

[@snackbob100]

Dude, this is fantastic! thank you. How can anyone dislike this i dont know!

[@11hamma]

he portrays lots of wrong info. non-beginners would know readily

[@LA-eq4mm]

@@11hamma like what

[@itjustmemyselfandi]

Can I ask how long it took to learn and watch this video?

[@odunayokomolafe9485]

Alright! This isi the most amazing tutorial I have seen with TensorFlow! I cant believe I can watch a tutorial for almost 7 hours being addicted. Thanks alot!

[@jsmammen6775]

Thank you for this video. This is the most thorough and simple introduction to Tensorflow and AI in general.

[@kawsydaisy]

Only 25 mins and already so good! Your videos never disappoint, Tim!

[@janicesmyth2183]

thank you so much Tim! I wish this was around when I was much younger! I was always very curious about learning about programming!

[@3T-InfoTinker]

Learning is something different than openionizing. Tim you are such a good teacher man.

[@guitarockdude]

Great Tutorial! Just a heads up, there was a mistake at 3:35:00 - you forgot to reinitialize the "model"!

[@enx1214]

New NN and tensorflow. I have searched and read lot before. Now I understood the different architecture, RNN vs CNN simple NN and they usage.

[@TemisBall]

Hmm, very deep. A y label called 'x' and an x label called 'y'. LOL, I really loved the video btw, I watched it until the end!

[@badboogl8529]

Yo, this tripped me up too lol P.S. 한국인이세요? 성험 때문에 물어요

[@tidtechnologyindepth6337]

I didn't understand that -1 thing at 54:25 , can anyone help me out!😭

[@AlenaShomanova]

@@tidtechnologyindepth6337 this is basically when you're telling to your machine "idk, I already gave you one number, count it yourself"

[@bengisu4592]

omg he is a very very good teacher. he explains everything in detail and very calmly

[@vierminus]

For those who stumble upon the error "AttributeError: module 'keras.preprocessing.sequence' has no attribute 'pad_sequences'" at 5:16:32: the pad_sequences function has been moved, you can do it like this now: train_data = keras.utils.pad_sequences(train_data, MAXLEN) test_data = keras.utils.pad_sequences(test_data, MAXLEN)

[@Jorvanius]

Dude, this course is amazing. I've only been through a third of it, but I know that I will watch it completely. Thanks you so much for sharing it

[@aparvkishnov4595]

I agree too coronavirus

[@Luxcium]

I love being taught by a cute and smart guy... This guy knows his topic has a clear bold voice and talks in a way that is easy to understand (I am not native English, I am from Quebec City)...

[@alexg2890]

3:37:01 Training on one epoch in this case builds on already existing model that was created using many epochs. You need to recreate the model to demonstrate this

I was thinking about this, and how tf he could actually get .9x in just one epoch

[@evanhagen7084]

Every video I've watched on Machine learning assumes we're in 1st grade and a college math major at the same time. "A tensor is a generalization of multidimensional vectors or matrices blah blah blah" 5 mins later "Slope is rise over run."

[@sangramjitchakraborty7845]

It's easier to explain what slope is then what a tensor is. It boggles your mind. Vector already represents multiple dimensions, and tensor is a multidimensional vector? That's pretty much impossible to visualise and get an intuition for. Slopes on the other hand, is rise over run.

I get what you mean, either you end up in some multivariable calculus explanaition or a lame ass "just a weighted sum bro aka Siraj Raval"

[@yasminamran5]

The best in the whole internet. This genius kid is amazing

[@rahulbhardwaj4568]

This is pure GOLD!!!!

[@gusinthecloud]

I had read 3 books about AI, before this video. You made a very clear course and helps me a lot. Thank you very much indeed.

[@abcdxx1059]

There are a lot of tutorials like this already available but there is less content about cleaning data or building pipelines it would be really helpful if you could make tutorials on it

[@RoboticusMusic]

Yep, or managing an automated 1D covnet with attention designed for time series prediction. This is both a desired topic at the moment and he could fundamentally demonstrate all the basics of building an automated pipeline where new data comes in, model updates and self optimizes, then outputs a prediction, process repeats upon completion or arrival of new data. This is something me and everyone getting into time series forecasting ML wants to see and it is not too ungodly complex like some other automated ML processes.

[@abcdxx1059]

@@RoboticusMusic all i am saying is that no one wants to make videos on the complex stuff almost all the content in this video has like 50 similar videos or blogs

[@RoboticusMusic]

@@abcdxx1059Very true. In the financial time series forecasting ML sphere I've only met one guy (Tom Starke) who has said anything rational. Everyone else is more or less Siraj. ML is not inherently bad but the industry is a huge elaborate scam like cryptocurrency. I haven't seen anyone build better ML models than the hand tuned hands-on real time manual adjustment algorithms I've built. I only need ML as icing on the cake to extract any last edge, and nobody seems to understand the basic principles of building a predictive model. For example none of the tutorials explain one-shot methods. That means everything out there overfits and it worse than useless! If a model can't learn in one episode it fundamentally is performing a very expensive database memorization hallucination.

[@MarcelinoSileoni]

Tim I think you've done a good job of introducing each of the topics you've touched on. I must say that in each topic, especially the latest and most complex ones, you have left many gaps to be covered by each of us. For the next video I recommend speaking more slowly, explaining in greater depth the fundamental concepts, the foundations that later serve to understand the practical examples. I also recommend preparing presentations instead of using a basic graphics application. However, congratulations for the courage to make the video without being an expert in the field.

[@ferozabraham9401]

Wonderful Job dear. God Bless!

[@justdevi]

Tim you're absolutely the best teacher i've ever had, thank you so much for doing this

[@bigdhav]

Tim is gonna be a CEO of a tech or education company in the future. What a legend.

[@aadam7459]

Just finished the entire video, your explanations were great and I got all the examples to work on my local machine, so kudos it was an amazing course :)

[@saiteja7170]

Another video need to be saved :) Thank you so much Tim!! ❤️

[@abujafarriju364]

মাওলানাআলতাফ

[@chhhhh2768]

Good intro, but hands down PyTorch > Tensorflow. Worked with Tensorflow and went to PyTorch and neverlooked back. Just easier to develop, test and explore and to know what you're actually doing. Pytorch almost feels like you're developing regular Python while Tensorflow feels like you are rearraning your Livingroom, but jou need to do it through a locked door via the keyhole with a long wire. Serving is a bit worse, but not that much and it improved.

[@kuravasic]

OMG dude you're lit. I've just watched all 7 hours, great course!

[@raghuveerjayanth2641]

im 14 and i was trying to learn tensorflow for like 2 years. You saved my life.

[@Zyger-xn3df]

you 15 now ?

[@brokenvectors]

46:55 don't mind me, just reminding myself

[@anirbanmalick7626]

Not sure if it is already discussed or not. At 3:37:00 you updated the model by running the same cell for multiple epochs. There the previous model got updated and thus accuracy improved. Not that, with less epoch the accuracy is high. Thanks :)

[@marcel4366]

Great tutorial and also great explanations! Thanks for that. Just having a remark, that you actually use tensorflow 2.1, but as you use a lot of tf.compat.v1 (eg for Session), this is more of a tensorflow 1.x-2.1 tutorial, as Sessions are not part of the official workflow anymore (as can be guessed by the ".compat" -> just for compatibility)

[@juliosouto9659]

Does tf.print(tensor) replace the session eval?

[@p.cnunes7098]

Free code camp is simply the best, hope to contribute to the project in the future, went from 0 to hero with your courses

[@Xarderrr]

What a timing! I've just finished a teoreticall ml course and it's time for some practise :D