Which of the followings is NOT correct about AI?
A system that acts like humans is not an AI system as it might generate wrong
Which statement is NOT correct about human neurons?
The output of neurons can be represented as a linear model
True or False: Deep learning is part of AI
True or False: Neural Network can be used as a supervised learning method
Is Stock value prediction an unsupervised learning process?
Which of the followings is NOT correct?
A neuron only takes one input and generates one output
Which of the following statement(s) correctly represent(s) a neuron?
A neuron can have a single input and a single output only
. A neuron can have multiple inputs but a single output only
A neuron can have a single input but multiple outputs
A neuron can have multiple inputs and multiple outputs
True or False: The larger the inputs are the better the learning process will be.
What is a neural network?
A mathematical model inspired by the structure and function of the brain
True or False: The gradient finds the new weights only based on the existing weights
True or False: If the inputs are large, the activation function gets very flat
What is the function of the activation function in a neural network?
To introduce non-linearity into the network
What is the primary goal of training a neural network?
To minimize the difference between the predicted and actual values on the
Which of the following is a common optimization algorithm used in neural network
What is backpropagation?
The process of updating the weights in the network during training
Which of the following is an example of unsupervised learning?
Given the target output and actual output, which one of the following functions is the
most common error function in neural networks?
(target – actual)^2
(Target - Actual)^2
Given the network outputs (0.4, 0.6, 0.2) and the target outputs of (0.3, 0.4, 0.1), what is to
network error value, if we use the square difference as error function:
True or False: The squared of difference is a common error function in neural network, because the error function is smooth and continuous making gradient descent work well ‐ there are no gaps or abrupt jumps
If there is a neural network classifying hand written numbers (0, 1, 2, 3, 4, 5, 6, 7, 8, 9), how many output neurons will it have?
For a neural network classifying images of 28 x 28 pixels, how many neurons should it have?
What is each run-through on a neural network called?
What is overfitting?
when the NN becomes too adjusted to the training set, causing error to be high on new data sets
If you have labeled data, would you want to use:
classification with supervised learning
If you have unlabled data, would you use:
clustering with unsupervised learning
Classification or Clustering: Credit Card Fraud Detection
Applications of AI in Finance
3 Main Areas of AI in Finance
Fraud detection and compliance, Banking chatbots and robo-advisory services, Algorithmic trading
What is Benfords Law
the principle that in any large, randomly produced set of natural numbers, most entries will begin with the digit 1, less begin with 2, and so on, with the smallest percentage beginning with 9
Triggers / Indicators for fraud
high amount transaction
many different locations in a short period of time
How does a NN learn
past experience (historical data)
What % of trades do computers make up?
50 - 70%
Benefits of Algorithmic Trading
types of algorithmic trading strategies
True or False: Signal processing is the algorithmic trading method where a mathematical extension of a technical analysis based on the art of filtering is used to eliminate
noise and discern trading patterns
T or F: the algorithmic trading strat "Pattern Recognition", the algorithm learns market psychology of supply and demand
What types of clinical activities are done to provide the data healthcare algorithms are trained on?
Which is the most used application of AI in healthcare
What is NLP (Natural Language Processing)
Extract information from unstructured data such as clinical notes/medical journals to supplement and enrich structured medical data
Commonly Used Deep Learning Algorithms in Medical Applications
Convolution neural network (CNN)
Recurrent neural network (RNN)
Deep belief network (DBN)
Deep neural network (DNN)
T of F: Convolutional neural network was first proposed and advocated for the high-dimensional image analysis
How accurate is the CNN on diagnosis and treatment selection
4 main components of each layer in CNN
Convolutional kernels defined by a width and height
number of input channels and output channels
depth of the convolution kernel/filter (equal the number channels of the input feature map.
hyperparameters of the convolution operation, like padding size and stride.
Two main components of NLP (natural language processing)
a challenge of learning models in AI
can only see the world as a repetition of the past
Applications of AI in Education
Data-based Neural AI, Logic-based AI, Knowledge-based AI
Transfer learning consists of:
learning new tasks that rely on previously learned tasks
T or F Reinforcement Learning consists of amplifying behavior that leads to outcomes that are defined as positive
Deep Learning Techniques in Cybersecurity
Deep Belief Networks,
Recurrent Neural Network
Convolutional Neural Networks
Generative Adversarial Networks
Recursive Neural Networks
T or F: Deep belief networks are trained in a supervised manner
T or F: Recurrent neural network processes inputs one element at a time, using the output of the hidden
units as additional input for the next element