Hands on ML Flashcards
What type of machine has the following characteristics:
-It takes a particular statistical model
-Optimizes it using a computer
-Leverages the computer speed
-Creates a class of models
a. Machine Learning
b. Deep Learning
c. AI
a. Machine Learning
-It takes a particular statistical model
-Optimizes it using a computer
-Leverages the computer speed
-Creates a class of models
What type of machine has the following characteristics:
-Deep neural networks
-It has cheaper computational power and can withstand a rise of alot of data and cheaper on cost of storage
a. Machine Learning
b. Deep Learning
c. AI
b. Deep Learning
-Deep neural networks
-It has cheaper computational power and can withstand a rise of alot of data and cheaper on cost of storage
Data is the key to _____
a. Machine Learning
b. Deep Learning
c. AI
a. Machine Learning
Data is the key to Machine Learning.
Without data, there is no machine learning.
What type of machine has the following characteristics:
-Learn data sets from a large set of unsupervised data
-Take this model and transfer this learning into a specific task
-Reduces data requirements of your target task
a. Machine Learning
b. Deep Learning
c. AI
d. Transfer Learning
d. Transfer Learning
-Learn data sets from a large set of unsupervised data
-Take this model and transfer this learning into a specific task
-Reduces data requirements of your target task
When programming a function, we take inputs or multiple inputs called ______
a. arguments
b. args
c. kwags
d. none
a. arguments
When programming a function, we take inputs or multiple inputs called arguments
What type of problems use Group-Based Programming?
a. image classification
b. self-driving cars
c. chess playing
d. game playing
e. all above
e. all above
Image classification, Self-driving cars, Chess Playing, and Game playing all use Group Based Programming.
Which type of ML model is the following:
“We take a dataset of input and output pairs and we get them to MLmodel to learn a function that transforms the input and outputs after it has seen enough input output pairs”
a. Reinforcement
b. Supervised
c. Unsupervised
d. Semi-Supervised
b. Supervised Learning
“We take a dataset of input and output pairs and we get them to MLmodel to learn a function that transforms the input and outputs after it has seen enough input output pairs”
Which type of ML model is the following:
“Holy Grail of ML, according to Deep Learning. We do not feed input-output pairs. We just feed inputs. Figures out the rules, patterns, and the underlying structure governing the data. Allows the machine to do all sorts of structured learning from unstructured data?”
a. Reinforcement
b. Supervised
c. Unsupervised
d. Semi-Supervised
c. Unsupervised
“Holy Grail of ML, according to Deep Learning. We do not feed input-output pairs. We just feed inputs. Figures out the rules, patterns, and the underlying structure governing the data. Allows the machine to do all sorts of structured learning from unstructured data?”
How data enables ML?
Using an ex.
input bank of pictures and output of labeled pictures of 1000 categories. Task: get the MLM to learn what pics go to what outputs.
Model tunes the own parameters to the dataset. Model turns into its own approximations. Parameters are the rules. Neural Network (model) and Weights(rules)
input bank of pictures and output of labeled pictures of 1000 categories. Task: get the MLM to learn what pics go to what outputs.
Model tunes the own parameters to the dataset. Model turns into its own approximations. Parameters are the rules. Neural Network (model) and Weights(rules)
What is the basic idea of a Feedback loop?
a. Compute the current prediction
b. Then look at how wrong you are
c. Then tune the model to reduce the error
d. All the above
The Basic idea of a Feedback Loop
d. All the above
-Compute the current prediction
-Then look at how wrong you are
-Then tune the model to reduce the error
What program is the following:
-“Provides high performance, easy to use data structures and data analysis tools for Python”
-“It is a data analysis library”
-“Works like Excel combines with SQL w/ very Pythonic interface
-“Import from Pandas then clean, slice, aggregate, or summarize”
a. Python
b. Pandas
c. Scikit Learn
d. Tensorflow
b. Pandas
-“Provides high performance, easy to use data structures and data analysis tools for Python”
-“It is a data analysis library”
-“Works like Excel combines with SQL w/ very Pythonic interface
-“Import from Pandas then clean, slice, aggregate, or summarize”
What program is the following:
-Very easy to go from code to output
-Run on public cloud services like AWS, Azure, and GCP
a. Python
b. Pandas
c. Scikit Learn
d. Tensorflow
e. Jupyter Notebook
e. Jupyter Notebook
-Very easy to go from code to output
-Run on public cloud services like AWS, Azure, and GCP
What program is the following:
-Patterns in data
-How values correlate to time
-How the value change in the future
a. Python
b. Time Series Data
c. Scikit Learn
d. Tensorflow
b. Time Series Data
-Patterns in data
-How values correlate to time
-How the value change in the future
pd.date_range(‘2019-01-01’ , periods=3, freq=’H’)
What does this mean?
pd.date_range(‘2019-01-01’ , periods=3, freq=’H’)
H - Hourly
3 - Means we want 3 periods
2019-01-01 is the date range the computer recognized
pd.date_range(‘2019-01-01’, periods=3, freq=’H’, tz_localize(‘UTC’)
What does this mean?
H = hourly
2019-01-01 = Date range starts
Periods = We want 3 periods
UTC = Time Zone
pd.date_range(‘2019-01-01’, periods=3, freq=’H’, tz_localize(‘UTC’) .tz_convert(‘US/Pacific’)
What does this mean?
H = hourly
2019-01-01 = Date range starts
Periods = We want 3 periods
Changed the time from UTC to US Pacific
What is the main matrix in linear regression?
a. mean_squared_error
b. min_squared_error
c. linear_regression
d. none of the above
a. mean_squared_error
mean_squared_error is one of the MAIN MATRIX in LINEAR REGRESSION
what function does the following:
“it is an error function that squares the errors of all the data points and divides them and finds the mean”
a. mean_squared_error
b. min_squared_error
c. linear_regression
d. none of the above
a. mean_squared_error
“it is an error function that squares the errors of all the data points and divides them and finds the mean”
how do you perform a train test split from scikitlearn?
a. from sklearn.model_selection import
train_test_split
b. test_train_split
c. train_test_split
d. none of the above
From scikitlearn you would perform train test split below:
a. from sklearn.model_selection import
train_test_split
what function does the following:
“turns any variable fit in between 0 and 1”
a. mean_squared_error
b. sigmoid (aka: logistic function)
c. linear_regression
d. none of the above
b. sigmoid (aka: logistic function)
“turns any variable fit in between 0 and 1”
ex on coding:
def sigmoid(z):
Math function
1/(1 + np.exp(-z))
What is the right percentage to Train and Test data?
a. Train 70%, Test 30%
b. Train 30% Test 70%
c. Train 40% Test 60%
d. Train 60% Test 40%
The right percentage to Train and Test data
a. Train 70%, Test 30%
What is another option other than a basic
“Train-Test” dataset?
a. Train-Test-Validate
b. Train-Test
c. Train-Test- Split
d. None of the Above
a. Train-Test-Validate
Train %
Test %
Validate %
(use validate to avoid spoiling data by giving too much data)
what function does the following:
“Forces a model to learn too much from the data”
“Model does not remember the data sets”
a. underfitting
b. overfitting
c. linear_regression
d. none of the above
b. overfitting
“Forces a model to learn too much from the data”
“Model does not remember the data sets”
How do you prevent Overfitting?
a. regularization
b. lasso
c. linear regression
d. underfitting
a. regularization
- do a train-test split
2.
takes all the data and scales it too an interval of 0 and 1.
a. polynomialFeature
b. standardscaler()
c. none of the above
b. standardscaler()
transforms the data into a 3rd degree polynomial feature set.
a. polynomialFeature
b. standardscaler()
c. none of the above
a. polynomialFeature
transforms the data into a 3rd degree polynomial feature set.
For this tactic the outcome is
“penalizes large weights. The higher the weight the more penalized it gets. The coef (coefficients) are more sensible. Ex. most coef. fall between 0 and 1”
A. Lasso
B. Ridge
C. ElasticNet (Lasso + Ridge)
B. Ridge
“penalizes large weights. The higher the weight the more penalized it gets. The coef (coefficients) are more sensible. Ex. most coef. fall between 0 and 1”
For this tactic the outcome is
“Minimize number of nonzero weights entirely.”
“Most coef. are 0”
A. Lasso
B. Ridge
C. ElasticNet (Lasso + Ridge)
A. Lasso
For this tactic you your outcome is
“Minimize number of nonzero weights entirely.”
“Most coef. are 0”
For this tactic the outcome is
“A type of tactic that is used Lasso + Ridge”
A. Lasso
B. Ridge
C. ElasticNet (Lasso + Ridge)
ElasticNet is a balance between Lasso and Ridge
What form of OpenAI is the following:
-Conversational Chatbot
-Generative Pre-trained Transformer
-Launched as prototype on 11/30/22
-Uses: Debuging/Writing programs, Compose Music, Write Essays, assignments, etc.
A. ChatGpt
B. Whisper
C. DALL-E
A. ChatGpt
-Conversational Chatbot
-Generative Pre-trained Transformer
-Launched as prototype on 11/30/22
-Uses: Debuging/Writing programs, Compose Music, Write Essays, assignments, etc.
What form of OpenAI is the following:
-Transcription and Translation
-Think of Google Translate on Steriods
A. ChatGpt
B. Whisper
C. DALL-E
B. Whisper
-Transcription and Translation
-Think of Google Translate on Steriods
What form of OpenAI is the following:
-Image Generation
A. ChatGpt
B. Whisper
C. DALL-E
-Image Generation
C. DALL-E
True or False:
AI, ML, and DL - Learn from past patterns to predict future
True
AI, ML, and DL - Learn from past patterns to predict future
What form of AI handles Structured data?
Think (Rows + Columns)
a. AI
b. ML
c. DL
b. ML
ML deals with Structured Data (such as Rows and Columns)
What form of AI handles Unstructured data?
Think (Images + Text + Videos + Emails)
a. AI
b. ML
c. DL
c. DL
DL handles Unstructured data
Think (Images + Text + Videos + Emails)
What are examples of Weak AI?
A. AI, ML, DL
B. AGI (Artificial General Intel.,), ASI (Art.Super Intel)
C. Supervised, Unsupervised, and Reinforcement
Weak AI
A. AI, ML, DL
What are examples of Strong AI?
A. AI, ML, DL
B. AGI (Artificial General Intel.,), ASI (Art.Super Intel)
C. Supervised, Unsupervised, and Reinforcement
Strong AI
B. AGI (Artificial General Intel.,), ASI (Art.Super Intel)
What form of AI is like Human Like Robots depicted in movies?
A. AI, ML, DL
B. AGI (Artificial General Intel.,), ASI (Art.Super Intel)
C. Supervised, Unsupervised, and Reinforcement
A form of AI is like Human Like Robots depicted in movies.
B. AGI (Artificial General Intel.,), ASI (Art.Super Intel)
ChatGPT touches the outer layer of Strong AI
“Basic form of DL, that consists of Input, Hidden, and Output layers?”
A. Artificial Neural Network (ANN)
B. Convolutional Neural Network (CNN)-Images
C. Recurrent Neural Network (RNN)-NLP
A. Artificial Neural Network (ANN)
Basic form of DL, that consists of Input, Hidden, and Output layers
“The more advanced form of DL, created to handle images and videos?”
A. Artificial Neural Network (ANN)
B. Convolutional Neural Network (CNN)-Images
C. Recurrent Neural Network (RNN)-NLP
B. Convolutional Neural Network (CNN)-Images
The more advanced form of DL, created to handle images and videos
“Helpful in Natural Language Processing, NLP used in Alexa and Siri,
-Good at predicting future
-Information goes through a cycle of loops
-Considers a current input and also what learned previous inputs
A. Artificial Neural Network (ANN)
B. Convolutional Neural Network (CNN)-Images
C. Recurrent Neural Network (RNN)-NLP
C. Recurrent Neural Network (RNN)-NLP
-Helpful in Natural Language Processing, NLP used in Alexa and Siri
-Good at predicting future
-Information goes through a cycle of loops
-Considers a current input and also what learned previous inputs
-Good for extended sequences
-Use self-attention mechanisms to weigh the importance of different parts of the input at each position
-Allow them to process sequential data
A. Artificial Neural Network (ANN)
B. Convolutional Neural Network (CNN)-Images
C. Recurrent Neural Network (RNN)-NLP
D. Transformers
D. Transformers
-Good for extended sequences
-Use self-attention mechanisms to weigh the importance of different parts of the input at each position
-Allow them to process sequential data
“Is a transformer based language model”
“Designed for language generation tasks (Translation, summarization and text completion)
“Based on deep neural architecture”
“Self attention mechanism- has relative weights for different parts of input & effectively process sequential data”
A. ChatGpt
B. Whisper
C. DALL-E
A. ChatGpt
“Is a transformer based language model”
“Designed for language generation tasks (Translation, summarization and text completion)
“Based on deep neural architecture”
“Self attention mechanism- has relative weights for different parts of input & effectively process sequential data”
What are the requirements for App Development in AI
- Tech Stack (JS, HTML/ CSS, AI ML (most impt)
stringlit takes care of all these requirements - Database
- Backend
- User Interface
- Hosting
- Tech Stack (JS, HTML/ CSS, AI ML (most impt)
stringlit takes care of all these requirements - Database
- Backend
- User Interface
- Hosting
what is data about data?
a. metadata
b. data
c. metaanalysis
d. none of the above
a. metadata
metadata- data about data
True or False: A 1/3 of Whisper’s audio audience is not english.
True!
1/3 of Whisper’s audio audience is not english
What platform is used to detect AI generated text?
a. ChatGPT
b. GPTZero
c. AI capture
b. GPTZero
GPTZero is used to detect AI generated text
“A measure of how well a probabilty model predicts a sample”
a. preplexity
b. impreceptiliby
c. language model
a. preplexity
“A measure of how well a probabilty model predicts a sample”
Is the probability distribution over sentences or sequence of words
-Evaluate the goodness of already written sentences
a. Perplexity
b. language model
c. natural language processing
b. language model
Is the probability distribution over sentences or sequence of words
-Evaluate the goodness of already written sentences
What are limitation of ChatGpt
a. Cant access internet
b. accuracy issues
c. limited by data 2021
d. all above are limitation
d. all above are limitation
-Cant access internet
-accuracy issues
-limited by data 2021
What are employment risks of using ChatGPT?
a. affect the demand for knowledge workers
b. lower quality of work
employment risks of using ChatGPT
a. affect the demand for knowledge workers
b. lower quality of work
What are cybersecurity risks of using ChatGPT?
a. create program to create viruses.
cybersecurity risks of using ChatGPT
a. create program to create viruses.
What are education risks of using ChatGPT?
a. statement of English and articles are dead. Papers dead because internet is writing for you.
education risks of using ChatGPT
a. statement of English and articles are dead. Papers dead because internet is writing for you.
