1. How you can define Machine Learning?
Where the Machine(Model) trying to learn the pattern of the dataset by applying the various algorithms and statistical models with respect to the dataset is known as Machine Learning
2. What do you understand Labelled training dataset?
The Training dataset which have the solution to the problem or task is called a Labelled Training dataset
3. What are 2 most common supervised ML tasks you have performed so far?
Regression and Classification
4. What kind of Machine learning algorithm would you used to walk robot in various unknown area?
By using Reinforcement Learning. Because Reinforcement learning can able to perceive and interpret its enviroinment, take actions and learn through trial and error.
5. What kind of ML algorithm you can use to segment your user into multiple groups?
Unsupervised ML algorithms i.e clustering algorithms are used to create multiple groups. Eg: K-mean clustering, DB scan, K-mean ++, Mini Batch k-mean, Hierarchical clustering
6. What type of learning algorithm realized on similarity measure to make a prediction?
Instance-based algorithm. This algorithm doesn’t generalize the model instead they compare the new problem to the training problem. whether the same type of problem is seen in the training or not. If seen based on this similarity it will do the prediction.
That’s why this algorithm is also called the lazy algorithm.
Eg: k Nearest Neighbour(KNN)
7. What is an online learning system?
In online learning, you can train the model incrementally by sending it data in sequential, individual or small groups(mini-batch) . Learning is fast and cheap. Online learning is great for the system that receive data in the continuous flow.
Online Learning can used to train system on huge dataset that cannot fit in machines main memory.
The big challenge in the online learning is if the bad data fed into the system then the system’s performance gradually decline.
8. What is out of core learning?
Out of core learning system is a system that can handle huge dataset but cannot fit in the machines main memory. In this situation we can online learning
9. Can you name a couple of ml challenges that you have faced?
A. Not enough training data
B. poor quality data (lots of outliers, errors)
B. Overfitting of the model
C. Underfitting of the model
D. Irrelevant features
10. Can you please give 1 example of hyperparameter tuning wrt some classification algorithm?
In knn performing Hyperparameter tuning along with GridSearchCV
'algorithm' : ['auto', 'ball_tree', 'kd_tree', 'brute'],
'leaf_size' : [10 , 15 , 20 , 25 , 30 , 35 , 45 , 50 ],
'p' : [1,2],
'weights' : ['uniform', 'distance'] }
grid_cv = GridSearchCV(knn,param_grid=pram