42 interview questions on Deep learning - Fundamentals for developers (2025)

Easy

15 votes760 runs0 comment

What is the missing word in the following sentence: Overfitting is the production of a model that corresponds too closely to the training data set and may therefore fail to fit other data sets and so fail to perform any reliable forecast.

Data Scientist

overfitting

Martinat Mar 14, 2022

Medium

14 votes737 runs0 comment

What is the relation between Artificial intelligence (AI), machine learning and deep learning?

Data Scientist

machine learning

deep learning

Martinat Sep 19, 2022

Easy

8 votes746 runs0 comment

Given a neural network with 4 inputs, how many connections could exist between the input layer and the first hidden layer?

Data Scientist

neural network

connections

Martinat Apr 4, 2022

Easy

4 votes688 runs0 comment

Which of the following processes are machine learning ones but not particularly related to deep learning?

Data Scientist

machine learning

random forest

Martinat Mar 14, 2022

Easy

1 vote717 runs0 comment

Which neural network type is the best for image processing?

Data Scientist

neural networks

image processing

Martinat Sep 19, 2022

Easy

1 vote701 runs0 comment

How to define unsupervised learning? In unsupervised learning ...

Data Scientist

machine learning

unsupervised learning

Martinat Mar 14, 2022

Easy

1 vote698 runs

2 comments

In deep learning, the validation set is used to provide frequent and unbiased evaluations of the model’s fit on the training set while tuning its hyperparameters/parameters: in other words,the model is found and then tested on the validation set before to be improved once again.

Data Scientist

hyperparameters

validation

Martinat Mar 14, 2022

Medium

1 vote673 runs1 comment

In order to get good results when implementing deep learning, one needs to prepare his data before using it. What action CAN be done during this preparation?

Data Scientist

data

preprocessing

Martinat Apr 4, 2022

Easy

1 vote670 runs0 comment

Cross validation is a technique used for evaluating DL (deep learning) models (an ML models in general). The method consists in training several models on subsets of the input data and evaluating them on the complementary subset of this same data. It aims at avoiding overfitting.

Data Scientist

overfitting

validation

Martinat Apr 4, 2022

Medium

1 vote668 runs0 comment

What is/are the purpose(s) of a neuron’s activation function?

Data Scientist

activation function

non-linearity

Martinat Apr 4, 2022

Medium

1 vote123 runs0 comment

The learning rate is the variable which determines the gradient descent algorithm steps’ size. It is used when studying the cost function. How is this value chosen? What are the risks?

Data Scientist

gradient descent

cost function

Martinat Apr 4, 2022

Medium

1 vote66 runs1 comment

How many types of gradient descent algorithms are there?

Data Scientist

gradient descent

optimization

Martinat Mar 14, 2022

Medium

1 vote61 runs0 comment

Calculate the output of a neuron with 3 inputs and 3 weights.

Data Scientist

activation function

linear function

Martinat Mar 14, 2022

Easy

1 vote52 runs0 comment

How could we shortly define deep learning ?

Data Scientist

machine learning

neural networks

Martinat Mar 14, 2022

Easy

0 vote695 runs1 comment

What type of learning are the following task included?

Data Scientist

supervised learning

classification

Martinat Mar 14, 2022

Medium

0 vote693 runs0 comment

What is a weight in a neuron? What is its aim?

Data Scientist

neural network

weights

Martinat Mar 14, 2022

Hard

0 vote681 runs1 comment

What are the differences/similarities between a loss function, an error function, and a cost function?

Data Scientist

loss function

error function

cost function

Martinat Mar 14, 2022

Easy

0 vote679 runs0 comment

What is the aim of data normalization? Its aim is to transform data sets so that all of its values are in a given range (ex: 0 to 1) while keeping the model's statistical properties.

Data Scientist

data

normalization

Martinat Mar 14, 2022

Easy

0 vote677 runs0 comment

What is the difference between shallow NN (neural networks) and deep NN (neural network)?

Data Scientist

neural networks

deep learning

Martinat Mar 14, 2022

Easy

0 vote667 runs0 comment

What are the parameters that are set before the learning process begins in deep learning?

Data Scientist

parameters

neural network

Martinat Mar 14, 2022

42 interview questions on Deep learning - Fundamentalsfor developers (2025)

Test your technical skills before an interview of Deep learning - Fundamentals with 42 corrected questions. Beginner, Intermediate or Expert level.

What is the missing word in the following sentence: Overfitting is the production of a model that corresponds too closely to the training data set and may therefore fail to fit other data sets and so fail to perform any reliable forecast.

What is the relation between Artificial intelligence (AI), machine learning and deep learning?

Given a neural network with 4 inputs, how many connections could exist between the input layer and the first hidden layer?

Which of the following processes are machine learning ones but not particularly related to deep learning?

Which neural network type is the best for image processing?

How to define unsupervised learning? In unsupervised learning ...

In deep learning, the validation set is used to provide frequent and unbiased evaluations of the model’s fit on the training set while tuning its hyperparameters/parameters: in other words,the model is found and then tested on the validation set before to be improved once again.

In order to get good results when implementing deep learning, one needs to prepare his data before using it. What action CAN be done during this preparation?

Cross validation is a technique used for evaluating DL (deep learning) models (an ML models in general). The method consists in training several models on subsets of the input data and evaluating them on the complementary subset of this same data. It aims at avoiding overfitting.

What is/are the purpose(s) of a neuron’s activation function?

The learning rate is the variable which determines the gradient descent algorithm steps’ size. It is used when studying the cost function. How is this value chosen? What are the risks?

How many types of gradient descent algorithms are there?

Calculate the output of a neuron with 3 inputs and 3 weights.

How could we shortly define deep learning ?

What type of learning are the following task included?

What is a weight in a neuron? What is its aim?

What are the differences/similarities between a loss function, an error function, and a cost function?

What is the aim of data normalization? Its aim is to transform data sets so that all of its values are in a given range (ex: 0 to 1) while keeping the model's statistical properties.

What is the difference between shallow NN (neural networks) and deep NN (neural network)?

What are the parameters that are set before the learning process begins in deep learning?

42 interview questions on Deep learning - Fundamentalsfor developers (2025)

Test your technical skills before an interview of Deep learning - Fundamentals with 42 corrected questions. Beginner, Intermediate or Expert level.

What is the missing word in the following sentence: Overfitting is the production of a model that corresponds too closely to the training data set and may therefore fail to fit other data sets and so fail to perform any reliable forecast.

What is the relation between Artificial intelligence (AI), machine learning and deep learning?

Given a neural network with 4 inputs, how many connections could exist between the input layer and the first hidden layer?

Which of the following processes are machine learning ones but not particularly related to deep learning?

Which neural network type is the best for image processing?

How to define unsupervised learning? In unsupervised learning ...

In deep learning, the validation set is used to provide frequent and unbiased evaluations of the model’s fit on the training set while tuning its hyperparameters/parameters: in other words,the model is found and then tested on the validation set before to be improved once again.

In order to get good results when implementing deep learning, one needs to prepare his data before using it. What action **CAN** be done during this preparation?

Cross validation is a technique used for evaluating DL (deep learning) models (an ML models in general). The method consists in training several models on subsets of the input data and evaluating them on the complementary subset of this same data. It aims at avoiding overfitting.

What is/are the purpose(s) of a neuron’s activation function?

The learning rate is the variable which determines the gradient descent algorithm steps’ size. It is used when studying the cost function. How is this value chosen? What are the risks?

How many types of gradient descent algorithms are there?

Calculate the output of a neuron with 3 inputs and 3 weights.

How could we shortly define deep learning ?

What type of learning are the following task included?

What is a weight in a neuron? What is its aim?

What are the differences/similarities between a loss function, an error function, and a cost function?

What is the aim of data normalization? Its aim is to transform data sets so that all of its values are in a given range (ex: 0 to 1) while keeping the model's statistical properties.

What is the difference between **shallow** NN (neural networks) and **deep** NN (neural network)?

What are the parameters that are set before the learning process begins in deep learning?

In order to get good results when implementing deep learning, one needs to prepare his data before using it. What action CAN be done during this preparation?

What is the difference between shallow NN (neural networks) and deep NN (neural network)?