If you found this video valuable, then hit the *_like_* button👍, and don't forget to *_subscribe_* ▶ to my channel as I upload a new Machine Learning Tutorial every week.
Very few people explain things mathematically and a very few people want a mathematical explanation. People just want to code without understanding the algorithm. You and your subscribers are the best :)
you sir are a legend. I took several tutorials on machine learning , your videos are the only one that make sense to me. I don't know if you have any paid course out there, if you do please let me know, i will definitely purchase it. good luck :)
Thanks for the video Jay 💙 Just a simple question, at 5:31: You used the method (reshape) to modify Y but the (transpose) to modify X! Why don't we use transpose for both? I tried it and I think it works, otherwise you have other reason! Thanks again for your amazing content 😄
You can perform the operation using reshape or transpose. Both are fine. There is no specific reason for me to use reshape instead of transpose. You can use any 😇
hi, I had been learning machine learning by my own and had seen many videos, your explanation was remarkable, keep going, there is a clarity after we listen to your videos, great great, all the best for making more videos on all algos of ML
@@CodingLane certainly yes, kindly upload more videos which can teach us from the scratch so that it will be easy for us to understand better than blindly using the python machine learning libraries. Great job, keep going
well , in your cost function video you told that dCost/dW = (A-Y).X but in code you wrote that dCost/dW = (1/m)(A-Y).X should I multiply (1/m) or not?please tell me bro
thanks for this useful video .. I just have one question : I have a dataset for students performance in a course and I am required to split my dataset into 70% for training and 30% for testing without using sklearn .. How to do so?
Hey, thanks a lot for the video! So I'm facing a major problem. When I run the model, I am getting cost as NaN for every iteration after the 0th iteration. Why is this happening? How do I fix this? For context, I am using a different dataset (adult census income dataset from Kaggle) but all the preprocessing has been done and all the columns have numerical values.
Its because you might be taking very large “learning_rate” Try to reduce its value by 100 times or 10000 times or may be more. Once you see cost function takes some value which is not NaN, you can increase the learning_rate or adjust it to train the model faster. If still it shows NaN, then check if you have implemented the equations of logistic regression properly or not. A slight change in equation can also cause model not to train.
Hi At 1:11 you are uploading csv files for train and test. I am using Google Colab. Thus, the code to upload the files I got was =files.upload() Thus, how do I fit the same using Pandas as demonstrated by you?
Hello, here are the ways to use the files on google colab and load into pandas: towardsdatascience.com/3-ways-to-load-csv-files-into-colab-7c14fcbdcb92 Hope it helps!
Sure. Let say if A = [0.2, 0.7, 0.8, 0.3, 0.4, 0.6] Then A > 0.5 will be = [false, true, true, false, false, true] And if you convert it into integer, then it will be, Afinal = [0, 1, 1, 0, 0, 1] Thus, A initially were just probabilities. Now Afinal are predictions for class 0 and 1
@@CodingLane hey thanks but accuracy is sum of all the correct predictions / total predictions meaning and apologies if I am wrong ... compared to y truth and y predict how many in [0, 1, 1, 0, 0, 1] were right / total , not simply > 0.5 which yes will simply separate the classes ... i did calculate it back yesterday and my accuracy was around 68 - 71 % .. Super sorry if i did it all wrong and big thanks again
def accuracy_manual(slope, intercept, X_test,Y_test): predictions = np.dot(slope.T, X_test) + intercept predictions_log =(sigmoid(predictions)) all_predictions=[1 if i >= 0.5 else 0 for i in predictions_log[0]] print("all predictions == ", len(all_predictions)) count=0 for i in ap: if Y_test[0][i] == all_predictions[i]: count+=1 else: pass print("correct count ", count) #alternate way s= sum([all_predictions[i] == Y_test[0][i] for i in all_predictions]) print("correct count ", s) #accuracy = correct count / total count accuracy = count/len(all_predictions) print("accuracy of model ", accuracy) return accuracy
Thanks a lot. your explanation was just awesome. Would you please make a similar video on Multiclass Logistic Regression from scratch? I am expecting it from you bro.
