TensorFlow Binary & Multi-class Classification

[@derekbanas]

MY UDEMY COURSES ARE 87.5% OFF TIL May 23rd ($9.99) www.udemy.com/ ➡ Python Data Science Series for $9.99 : Highest Rated & Largest Python Udemy Course + 56 Hrs + 200 Videos + Data Science bit.ly/Master_Python_31 ➡ New C++ Programming Bootcamp Series for $9.99 : Over 23 Hrs + 53 Videos + Quizzes + Graded Assignments + New Videos Every Month bit.ly/C_Course_31

[@ChessFlix]

Thanks for continuing to make videos Derek! I'm not even particularly interested in Tensorflow right now, but I just enjoy your material. If you're taking requests for the future, I'd love to see you cover some js/ts stuff with a frontend framework. React, Vue, Svelte, or Solid in TypeScript would be super entertaining.

[@derekbanas]

Thank you for taking the time to write a nice note :) I’ll see what I can do about your requests

[@ianhaylock7409]

Hi, it would be nice if you could explain the differences between the different types of neural networks, and which type would be best for which type of data set.

[@derekbanas]

After I cover everything in detail I will then make a video that pulls all the concepts together as you requested. Thank you for the input :)

[@hakanbozcuk7761]

Thanks for your valuable instruction. In the binary classification part, how can you calculate the outcome(good or bad wine) for an unknown wine for which you have the features, buy using model_1?

[@jordansocha6932]

1. In the binary classification dataset the target variable (quality) was distributed as 1,271 rated as "good wine" and 5,192 rated as "not good wine" for a total of 6,463 samples. Since the output is heavily distributed toward "not good wine" will the model pickup on this bias and lean toward predicting "not good wine?" Does the dataset need to be "evened out" so the number of "good wine" and "not good wine" samples are the same (or at least roughly the same)? 2. I understand you set the random state to 66 for reproducibility; if you did not do this and you use tf.random.shuffle and ran the model many times would this be similar to cross-validation?

[@timstevens3361]

nice job Derek !

[@derekbanas]

Thank you for taking the time to tell me you liked it :)

[@martinndungu7803]

This was awesome💪 looking forward to you doing NLP with tensorflow

[@derekbanas]

Thank you :) I'll be covering NLP very soon

[@andrealcantara1437]

Shouldn't the evaluate be used with the X_test and y_test instead of X_train and y_train?

[@broooth]

Hey Derek, you evaluated your model on the training data, not testing (59:33)

[@derekbanas]

Sorry about that. I was shuffling a lot of things

[@canozturk8993]

Hey Derek, can you please make a video about how to boost our confidence and social courage ? Thanks in advance.

[@derekbanas]

I used to train sales people. I cover the science of how you can get people to like you here kzbin.info/www/bejne/bJOxqHWAacmer5o I have tons of scientifically tested ways of influencing people and such. Aside from that I have found the easiest thing to do is to entertain people. People love stories. When I'm in a social interaction I do my best to quickly shift into telling stories and then tell them. We all have interesting stories. Stories that are funny or self deprecating work the best. When I was dating when I was younger I used to always carry a deck of cards and would show girls card tricks. Any entertainment works great. It can work as simple as walking up to a stranger and saying hey I discovered this neat card trick. I'm bored. Would you like to see it? Be an entertainer and everyone will love you and other bored people will be drawn to you. I hope that helps.

[@bhavinmoriya9216]

Thank yo very much for awesome videos as always. Don't we have to set sigmoid as activation in output layer, while we do binary classification?

[@derekbanas]

Thank you for taking the time to tell me you enjoyed it :) Yes Sigmoid is used for output with Binary. Sorry if I didn't do that in the video for some reason. It is correct on Github.

[@andrealcantara1437]

Why did you used ct.fit only on X_train? Shouldn't be a fit with the train and other with the test data?

[@andrealcantara1437]

I've tested doing the fit_transform before the splitting and got the same result, but I don't understand why would you only need to fit the X_train, not the X_test too.

[@derekbanas]

Training data is used to prepare the model to guess accurately with the test data. Training Phase : I say this is a dog while showing a picture of a dog, or this is a cat while showing a picture of a cat Testing Phase : I show a picture of a dog and ask them, is this a dog or a cat? Does that make sense now?

[@morthim]

1:04:16 'what does correlation mean'?

[@derekbanas]

Correlation is when two things move in a similar way