Today we will talk about an important type of neural network, which is Convolutional Neural Networks or CNN. CNN plays a crucial role in computer vision. Computer vision as the name suggests aims at helping computers decrypt visual data and over time learn to recognise images.
In this process, CNN helps the computer in breaking down the image into pixels, perform mathematical operations on it and make predictions on what it is seeing. This is widely used in applications such as image and video recognition, medical imaging, object detection, and facial recognition.
To employ these technologies and build the right products, companies are actively looking for Data Scientists with knowledge of Deep Learning and experience in CNN.
If you are interviewing for a Data Scientist position that involves deep learning and artificial intelligence-related tasks, you should expect some questions on CNN in your interview. In this article, we have prepared a list of the top 10 Data Scientist interview questions on CNN. These questions will help you practice and prepare for your next Data Scientist interview.
Top 10 Convolutional Neural Network Interview Questions and Answers
Here is a list of top ten commonly asked convolutional neural network interview questions and answers in machine learning job interviews –
1. What are activation functions?
Activation functions are the functions that are used to evaluate the output of a node in a neural network. It is the sum of two quantities- the weighted sum of the input values and a bias term. The output value decides whether a particular neuron in a specific neural network layer is activated.
2. What are the different types of pooling layers in a CNN architecture?
Assuming the filter of size 2×2, the three common ways of creating a pooling layer in a neural network are:
- Max Pooling: It returns the maximum value out of the 4 elements in the receptive field.
- Average Pooling: In this case, the average of the four values is computed to produce the output.
- L2 Pooling: The square root of the sum of the squares of the four elements is computed.
3. What is padding? Why do we need padding?
Padding refers to the process of adding a layer of zeros/ones to the sides of a matrix. We perform padding so that we can smoothly implement the convolution operation.
4. What is transfer learning?
Transfer Learning is the process of utilizing the knowledge gained by the pre-trained model and its knowledge to solve a more complex problem.
5. Evaluate the size of a feature map, given that the image size is 32×32, filter size is 5×5, stride is 1, and no padding.
The size of the feature map is 28×28 as there is no padding.
6. What is a flattening layer in CNN architecture?
The flattening layer is usually towards the end of the CNN architecture, and it is used to transform all the two-dimensional matrices into a single lengthy vector. The output of this layer is passed to the fully-connected layer.
7. What is convolution?
Convolution is a mathematical operation performed to overlay the contents of one function over the other. Mathematically, the convolution of two functions, w and f, is given by
Where a = (m-1)/2; m denotes the size of w.
8. List a few applications of a CNN.
A few applications of CNN are:
- Face Detection and Recognition
- Image Classification
- Image Segmentation
- Edge detection
- Document Analysis
- Image Colorization
- Recommendation Systems
9. What do you understand by Stride?
Stride refers to the number of steps the filter matrix can shift after evaluating the convolution between input and filter. When the stride=1, the filter matrix shifts by one pixel; if it is 2, then the filter matrix must shift by two pixels.
10. What are the advantages of using CNN over ANN?
The primary advantage of using CNN over ANN are
-
A CNN architecture can detect the important features in an image on its own and does not require human intervention.
-
A CNN can solve classification problems with higher accuracy as compared to ANN.
Besides exploring these interview questions and answers, you must work on a few industry-level projects as it will help you better understand the applications of CNN in solving real-world problems.
Before your next Data Scientist and AI interview, we suggest you practice these questions and prepare answers. To practice more, check out our article on Top Machine Learning interview questions.
If you are new to Data Science, check out our Data Science course which helps you become a world-class Data Scientist in just 10-months. To know more, connect with us here.
