Machine Learning Interview Questions

Machine Learning Interview Questions:

1. What is Machine Learning (ML)?

We will begin with the most informal and simple definition. Machine learning is a thing-labeler where you explain your task with examples instead of instructions.

The concept behind Machine Learning is that you feed data to machines and let them learn on their own without any human intervention (in the process of learning). Consider a small scenario. Let’s say that you have enrolled for some swimming classes and you have no prior experience of swimming.

Needless to say that initially, you would perform not so well because you have no idea about how to swim, but as you observe and pick up more information, your performance keeps getting better.

The observance is just another way of data collection. Just like how we humans learn from our observations and experiences, machines are also capable of learning on their own when they are fed a good amount of data.

This is precisely how we employ the concept of Machine Learning. It’s the process of getting machines to learn and improve from experience without being explicitly programmed automatically.

2. What is Data Science?

- Data Science is the field of study that combine mathematics, statistic, programming skills, data analysis, domain expertise and machine learning.

- To extract the meaningful information from structural and  unstructured data to solve business and real word problems.

3. What are the types of Machine Learning?

- Supervised Machine Leaning

- Unsupervised Machine Learning

- Reinforcement Machine Learning

4. How to save Keras Model?

- For Sequential and Functional API use 

model.save("model_name.h5")

After

model.fit([..])

Keras will use the HDF5 format to save both the model's architecture  and the values of all the model parameter for every layer. 

5. How to load Keras Model?

- For Sequential and Functional API use 

model=keras.model("model_name.h5")

6. What is ModelCheckpoint callback?

- ModelCheckpoint callback  saves checkpoints of your model at regular intervals during training, by default at the end of each epoch.

Save the model for each Epoch:

checkpoint_cb = keras.callbacks.ModelCheckpoint("my_keras_model.h5")

Save model for best validation accuracy:

keras.callbacks.ModelCheckpoint("my_keras_modelh5",save_best_only=True)

7. How many callbacks available in keras?

-  Base Callback class

-  ModelCheckpoint

-  TensorBoard

-  EarlyStopping

-  LearningRateScheduler

-  LambdaCallback

-  RemoteMonitor

8. What is TensorBoard?

- TensorBoard is a great interactive visualization tool that you can use to view the learning curves during training, compare learning curves between multiple runs, visualize the computation graph, analyze training statistics, view images generated by your model.

- It comes with TensorFlow.

9.  Name 5 popular activation funtions. Can you draw them?

- In Neural Network , We use below most popular activation functions:

    1. ReLU (Rectified Linear Unit) Function.

    2. Sigmoid (Logistic) Function.

    3. SoftMax Function.

    4. Step Function.

    5. Tanh (Hyper Tangent) Function.

10. How many Neurons do you need in the output layer if you want to classify email is spam or not?

-  To classify email is spam or not. You just need one neuron in the output layer of a neural network - for example indicating the probability that the email is spam.

- You would typically use the logistic activation function (Sigmoid) in the output layer when estimating a probability.

- It gives the value 1( spam if target value is 1) or 0 ( Not spam if the target value is 0).

11. In which domain you can use Machine Learning?

- Banking

- Insurance

- Education

- Gaming

- Logistics

- Entertainment

12.  Tell me the applications of Machine Learning?

-  Using  Machine Learning we can build thousands of applications, like

    1. Credit-card fraud detection

    2. Face Recognition.

    3. Recommendation System.

    4. Self-driving Cars

    5. Object Classification

13.  Tell me 4 most important Python libraries to solve Machine Learning Project?

-    Numpy : Numerical  Calculation.

-    Pandas : Data Analysis.

-    Matplotlib : Data Visualization.

-    Scikit-Learn : Data Pre-processing and ML Algorithms.

14. What is Supervised Learning?