- Видео 136
- Просмотров 25 026
Practical Data Science and Machine learning
Индия
Добавлен 9 ноя 2011
I am a Data Scientist with over 10 + years of experience in IT industry on various domains. My expertise is in Python, Machine Learning and Deep learning. This channel is to help beginners to have an understanding of what is datascience, Machine learning and Artificial Intelligence and get started with it.
Efficient way to learn Maths for Machine learning | How to learn the MATH for Machine Learning?
In this video, I have tried to convey a message to all fellow machine learning and deep learning learners on how to learn math required for ML and DL without getting overwhelmed and bored. | How to learn maths for machine learning and deep learning efficiently?
Просмотров: 200
Видео
Out of Bag error/Out of Bag score (OOB Error/ OOB Score) |OOB Score/Error in Random Forest algorithm
Просмотров 1283 месяца назад
In this video, you will understand what is OOB score or OOB error in Random Forest machine learning algorithm
Adaboost Classification | Adaboost classifier | complete math explained
Просмотров 773 месяца назад
In this video, you will understand how adaboost machine learning algorithm works in detail with complete math and end to end explaination.
Bagging Vs Boosting | Difference between Bagging and Boosting Machine learning Algorithms
Просмотров 1783 месяца назад
In this video you will understand the main differences between bagging and boosting based on how they are trained, what kind of base learners used, and how the predictions are made on the trained models. In interviews Question can also be asked in a different way like: " What is the difference between random forest and Adaboost?"
Probability Vs Likelihood | What is the difference between probability and likelihood?
Просмотров 1674 месяца назад
In this video, you will understand the difference between likelihood and probability. Often confusing and asked in interviews also.
Understanding derivatives of activation functions in detail
Просмотров 654 месяца назад
In this video, you will understand how the derivatives of activation functions affects the neural network learning and why we need the values at each layer to be in small range while we train our neural networks. sigmoid derivative will be between 0 and 0.25 (inclusive) Tanh derivative will be between 0 and 1 (inclusive) Link to notebook to plot derivatives of activation functions: github.com/S...
Neural Network activation function Derivatives : Equations for derivatives of activation function
Просмотров 1204 месяца назад
watching this video will help you in understanding how we can differentiate sigmoid, tanh, relu and leaky relu activation functions. This will be important for you to understand and implement neural network from scratch. Derivatives of Activation functions. Derivative of sigmoid, Derivative of Tanh, Derivative of ReLu, Derivative of LeakyReLu
Implement Neural Network from scratch: Forward & backward pass in python | No keras, No Tensorflow
Просмотров 1914 месяца назад
In this video, you will learn how to implement a nerual network from scratch without using keras, tensorflow or pytorch library. It is a pure python implementation. GitHub link to access the code: github.com/ShankarPendse/DeepLearning-Tutorials/blob/main/neuralnetwork_from_scratch.py
Equations of Gradients of a Neural network: Derivations of gradients/derivatives of a neural network
Просмотров 1044 месяца назад
In this video, I have explained in detail how to derive the gradients/derivatives of the cost function with respect to weights and biases at different layers. I have considered binary cross entropy as cost function the one we use for binary classification. I have derived each and every gradient with complete math.
Probability Distributions - Part 1 : A Gentle introduction to Probability distributions
Просмотров 304 месяца назад
In this video, I am introducing you to what do we mean by Probability distributions , different forms it can take and based on the type of random variable we have Probability mass function and Probability density function.
Machine learning Interview Questions: Why Multicollinearity is a problem in Machine learning?
Просмотров 314 месяца назад
In this video, you will understand what is multicollinearity, why it is bad, how to solve it and which algorithms it affects.
Implement Custom Ensemble Machine learning model | Naive Bayes+Logistic regression + SVM
Просмотров 964 месяца назад
In this video, you will learn how to implement a custom ensemble machine learning model for classification task. Models used are: 1. Multinomial Naive bayes 2. Logistic Regression 3. Support vector machine. In the end, all those 3 models are ocmbined to give us overall a better model.
part 3 - Implement Forward pass of Neural Network from scratch in python
Просмотров 554 месяца назад
Implement Neural Network from scratch in python. In this video, you will learn how we can implement a forward propagation of a neural network from scratch in python just making use of numpy
Part 2 - Implement single layer of a Neural Network from scratch in python
Просмотров 614 месяца назад
Implement Neural Network from scratch in python. In this video, you will learn how we can implement two neurons in python from scratch and how you can generalize this idea for any number of neurons
Part 1 - Implement single Neuron of a Neural Network from scratch in python
Просмотров 694 месяца назад
Implement Neural Network from scratch in python. In this video, you will learn how we can implement a single neuron in python from scratch.
SVM: Support Vector Machine - Soft Margin | Complete math behind soft marging SVM
Просмотров 394 месяца назад
SVM: Support Vector Machine - Soft Margin | Complete math behind soft marging SVM
SVM: Support Vector Machine - Hard Margin | Complete math behind SVM hard margin classifier
Просмотров 674 месяца назад
SVM: Support Vector Machine - Hard Margin | Complete math behind SVM hard margin classifier
Support Vector Machine (SVM) - Geometric intuition
Просмотров 844 месяца назад
Support Vector Machine (SVM) - Geometric intuition
ElasticNet Regression | Machine learning Linear regression
Просмотров 235 месяцев назад
ElasticNet Regression | Machine learning Linear regression
SVM Pre requisite: What is Hyperplane? | Hyperplane equation in N-Dimensional Space
Просмотров 1695 месяцев назад
SVM Pre requisite: What is Hyperplane? | Hyperplane equation in N-Dimensional Space
Sklearn Pipelines in Machine learning | Column Transformers
Просмотров 1035 месяцев назад
Sklearn Pipelines in Machine learning | Column Transformers
Outliers detection using IQR and Z_Score | Machine learning | Data Science
Просмотров 305 месяцев назад
Outliers detection using IQR and Z_Score | Machine learning | Data Science
Boosting in Machine learning - Intuition | Ensemble technique
Просмотров 1705 месяцев назад
Boosting in Machine learning - Intuition | Ensemble technique
Random Forest | Machine learning Ensemble | Bagging | Bootstrap aggregation
Просмотров 1325 месяцев назад
Random Forest | Machine learning Ensemble | Bagging | Bootstrap aggregation
Ensemble Algorithms in Machine learning | Bagging | Boosting
Просмотров 265 месяцев назад
Ensemble Algorithms in Machine learning | Bagging | Boosting
Bagging in Machine learning | Explained in depth
Просмотров 365 месяцев назад
Bagging in Machine learning | Explained in depth
Machine learning mini project: Twitter Bot classification | Machine learning project
Просмотров 3805 месяцев назад
Machine learning mini project: Twitter Bot classification | Machine learning project
Machine learning Interview Questions: What if parameters of logistic regression are set to zero
Просмотров 295 месяцев назад
Machine learning Interview Questions: What if parameters of logistic regression are set to zero
Machine learning Interview Questions: Why logistic regression is called regression?
Просмотров 865 месяцев назад
Machine learning Interview Questions: Why logistic regression is called regression?
Steps to follow in any Machine learning / Data Science project
Просмотров 1645 месяцев назад
Steps to follow in any Machine learning / Data Science project
good video👍
Thanks you very munch
Thanks for the in-depth explanation.Very Helpful
I cant believe such good content has only 50 views, great video sir, helped me to understand this topic better
What’s gradient descent
It is one of the optimisation algorithms we use to minimize the overall loss(cost)
bro you were a great help keep making content
sir please give practical implemnetation if required otherwise where practical implemntaion is ncessary in GEN AI using GRU please explain
The models that we are seeing as LLMs make use of transformer architecture. LSTM, GRU, Attention models act as a foundation to have proper understanding of transformer modles
really great calculation thank you
Do you need to normalise the data?
Yes, For KNN regression or classification, normalizing the data is very important, all the considered features should be along the same scale. if the feature values vary largely, the features with large values will have more impact while calculating the distance and eventually will affect the prediction. In this video, if you look at the dataset, the feature values are along the same scale, they are not varying largely.. so I did not do normalization.
@@shankarpendse Thank you! My data does vary, so I will be sure to normalize it - this video is very helpful. I had also been wondering - for each entry in my dataset, should I create a new column for a 'Euclidean Norm' value? that I would use to compare values via Euclidean Distance?
Glad to hear that it is helpful. I do not think you need to save the distance. Think of it this way, if you want to save the distance for each data point with every other data points, you would need to save 'n-1' distances for each of the 'n' data points. So, you will end up with a matrix of nx(n-1) just to store the distances. Calculation itself would take time if n is large. So, I would not be saving the distances as another column. But yes, if you want to have a look at the values and then want to make use of this distances data matrix as input to your TSP (Travelling salesman kind of problems) then go ahead and store them.
Good
Sir i need some clarity on the level of DSA that recruiters expect, hope you can throw some light on it. Basically how much DSA is expected for becoming a data scientist?
Medium to hard questions on leetcode and hard questions on sql
“Highly appreciated”
Nice explanation!
Thanks man, it helped me a lot
You are welcome
link to the code: github.com/ShankarPendse/DeepLearning-Tutorials/blob/main/forwardpass.py
link to the code: github.com/ShankarPendse/DeepLearning-Tutorials/blob/main/two_neurons.py
link to the code: github.com/ShankarPendse/DeepLearning-Tutorials/blob/main/single_neuron.py
I have messed up the last part of explaination in this video. Treat Non spam as 0 (negative class) and spam as 1 (as positive class) then we are actually reducing False negatives that is, our model (custom ensemble) is actually doing good job in identifying the spam mails.
Is this NLP related
Deep Learning Implementations: ruclips.net/p/PLUZxSPLab6f6t52E61UkP3jRZXhHxvDNR This playlist has projects related to NLP
Sir could you upload a video showing how we can train a neural network and how the weights are updated without using in-built python libraries. It would be really useful as I am doing my final year project based on implementation of neural networks.
The bit you are looking for is back propagation and I will upload the video soon. The way you update the weights using gradients, is actually dependent on the cost function that you are trying to minimize. I can help you with the way you have to update the weights if you can let me know your neural network's cost function.
I have release a video on how to compute gradients for neural networks to update weights and biases, I have derived the gradients step by step in detail, you can check out that video to understand how it is done.
Implementation from scratch where I have showed how we update the weights in python, without using any inbuilt Deep learning library. I have just used numpy to vectorize the process: ruclips.net/video/3UPWmn4FtDs/видео.html
How to be successful in finding ml roles while you're transitioning from another data career . Currently I'm on SQL and reports . I'm preparing for ml and data science on my own since the day I got my job in SQL . Still I'm unable to find a role that accept fresh experience persons
when we used Random Forest with default parameters, the accuracy on test data was 88% but after hyper parameter tuning it was 84% so how did it become better. Please explain
The model was overfitting with default parameters.. on the training set accuracy was nearly 100% and on the test set it was around 88%, so it's the classical example of a high variance model. So we have to fine tune it to reduce overfitting. You can refer to this video: ruclips.net/video/cnSI2FcXINM/видео.html
@@shankarpendse i understood that, my doubt is cant we achieve the same 88% accuracy after the hyper param tuning? is it normal to get little reduced while the model variance gets balanced o both of them could happen
Answer to your second question is Yes. To increase the model performance: you can try out other pre processing techniques and also check by making use of text column to extract the features and then use them to train the model. You can also try balancing the dataset using minority oversampling technique and test the performance.
Please explain the division of hourly data into categorical data formula
Great doing
Good sir g very informative project thanks🥰
Sir, Please Make a project by following all the steps you have shown in this video. Also, do the experimentation part for the better understanding for us. This will make me more confident in making project in real-world scenario as nobody has made such video till now on RUclips. Thank You 😊
You have it now, it will be live on 4th August @ 8AM IST. Link to the video: ruclips.net/video/0YCVkEiz9Rw/видео.html
finally someone who will give industry level knowledge for free. Hoping to learn projects implemented with fast api & deployment with CI/CD..
I will definitely cover everything, taking it one step at a time
log of odds and maximum likelihood: ruclips.net/video/_1kIlACzrew/видео.html
Link to notebook: github.com/ShankarPendse/ML_tutorials/blob/main/k-means%20clustering.ipynb Link to dataset: github.com/ShankarPendse/ML_tutorials/tree/main/datasets/clustering
Link to notebook: github.com/ShankarPendse/Natural-Language-Processing/blob/main/RNN%20Language%20model%20Jack%20and%20Jill%20poem%20learning.ipynb
Please find the link to notebook: github.com/ShankarPendse/Natural-Language-Processing/blob/main/Sentiment%20analysis%20model%20using%20LSTM.ipynb remember, You can try to improve the model by adding another data preprocessing step, if you see the data carefully, there are some repeated characters for example: [mamamama, waaaaaaaaaaaa , fuuuu****dd, hjdjjsiurjhjfhoipwplkjdhjdhhjhjdxdhueuhrjjjjjaaayyyyyyyyyyyi] which I think might be contributing in dragging down the model training, you can remove these kind of words using regular expression and include it as part of preprocessing. Also, you can try to normalize the input data along the features which might help in faster training.
Nice explanation mama!!
How is scenario B High variance? There is just 3% difference ebtween train test accuracy?
It's a general term if there is a difference between train and test metrics, we should be trying to mitigate that, if not we can say that there is unavoidable bias or unavoidable variance
ChatGPT 4o suggested this video for understanding log of odds n logistic regression
Awesome 👍
Very usefull sir thank you❤
Thanks, Very Clear Explanation. i usually share your videos to my peers who are starting out in ML.
I appreciate that!
I've never seen the gen_from_text thing before. interesting!
He uses pd.read_csv in another video..
Thank you.... very helpful content
Hello sir, Having few doubts. Pls share your mail id.I want to connect with you
pshankar1306@gmail.com, apologies for late notice
to calculate gradient, there are different formulas for both parameters. in the code above u have given same gradient value to both slope and intercept value
Thanks, I wanted to see how how gradient descent minimising cost without coding it my self. 🙂 this vid is perfect.
Assalamualaikum
Sir what is the filename in this video
Great
what is 'a' called as in minkowski distance metric?