microsoft/AI-For-Beginners

Public

mirrored fromhttps://github.com/microsoft/AI-For-BeginnersAvailable

CodeCommitsIssuesPull requestsActionsInsightsSecurity
c17ecaaf539eaad30af92e859d49623eb3803241

Branches

Tags

  • No tags available.
0Branches0Tags
Go to file
Add file
Code

Clone

HTTPS

Download ZIP

etc/quiz-src/questions-en.txt

639lines · modecode

1Lesson 1B Introduction to AI: Pre Quiz
2* A famous 19th century proto-computer engineer was
3- Charles Barkley
4+ Charles Babbage
5- Charles Darwin
6* Weak AI is a system designed to solve many tasks
7- True
8+ False
9* Chat bots are an example of truly intelligent systems
10- false, they are usually designed by a series of rules.
11- true, they are usually considered to be 'intelligent
12+ false, but they are increasingly able to pass Turing tests as they become more sophisticated.
13
14Lesson 1E Introduction to AI: Post-Quiz
15* A top-down approach to AI is a model of reasoning called
16- strategic reasoning
17+ symbolic reasoning
18- synergistic reasoning
19* A bottom-up approach to AI is based on neural networks
20+ True
21- False
22* The AI Winter occurred in this era
23- 1950s
24- 1960s
25+ 1970s
26
27Lesson 2B Knowledge Representation and Expert Systems: Pre-Quiz
28* The top-down approach to creating intelligent systems was based on:
29- knowledge seeking and reading
30+ knowledge representation and reasoning
31- knowledge reasoning and seeking
32* Knowledge is the same as information
33- True
34+ False
35* Knowledge is obtained by an:
36+ active learning process
37- passive learning process
38- both of these
39
40Lesson 2E Knowledge Representation and Expert Systems: Post-Quiz
41* The simplest method of knowledge representation is:
42+ algorithmic
43- symbolic
44- synergistic
45* Scenarios can represent complex situations that can unfold in time
46+ true
47- false
48* Forward inference starts with initial data and then:
49+ executes a reasoning loop
50- looks for a goal
51- starts over
52
53Lesson 3B Introduction to Neural Networks - Perceptron: Pre-Quiz
54* Early neural networks required
55+ manual weight adjusting
56- terabytes of data
57- special reasoning
58* A simple neuron is also called a 'threshold logic unit'
59+ true
60- false
61* A perceptron is a ___ type of model
62- multi-class classification
63- clustering
64+ binary classification
65
66Lesson 3E Introduction to Neural Networks - Perceptron: Post-Quiz
67* To train a perceptron, find a weights vector that results in the smallest ___.
68- size
69+ error
70- nodes
71* To minimize the function of weights, you can use gradient descent
72+ true
73- false
74* During gradient descent, each step updates the ___
75- learning rate
76+ weights
77- gradient
78
79Lesson 4B Neural Networks: Pre Quiz
80* The quality of prediction is measured by Loss function
81+ True
82- False
83* One layer network is capable of classifying ____
84- linearly joined classes
85+ linearly separable classes
86- single layers of classes
87* The method of training multi-layered perceptron is called ____
88+ back propagation
89- multiple propagation
90- front propagation
91
92Lesson 4E Neural Networks: Post Quiz
93* We use ____ for regression loss functions
94- absolute error
95- mean squared error
96+ all of the above
97* All but one is a type of classification loss function
98- 0-1 loss
99+ binary loss
100- logistic loss
101* Cross-entropy loss is a function that can calculate similarity between two arbitrary probability distributions
102+ True
103- False
104
105Lesson 5B Frameworks: Pre Quiz
106* Deep Neural Network training requires a lot of computations
107+ True
108- False
109* Overfitting occurs because of ____
110- Not enough testing data
111+ Too powerful model
112- Too much noise in output data
113* Bias errors are caused by our ____ not being able to capture the relationship between training data correctly.
114- model
115+ algorithm
116- computer
117
118Lesson 5E Frameworks: Post Quiz
119* After compiling our model object, we train by calling ____ function
120+ fit
121- train
122- teach
123* Binary cross-entropy is also called log loss
124+ True
125- False
126* TensorFlow is to ____ while PyTorch is to ____
127- Facebook, Google
128+ Google, Facebook
129- Microsoft, Google
130
131Lesson 6B Introduction to Computer Vision: Pre Quiz
132* Computer vision aims to allow computers gain understanding of _____
133+ images
134- text
135- computers
136* Python libraries available for image processing includes
137- OpenCV
138- Pillow
139+ a and b
140* Images cannot be represented as NumPy arrays in Python
141- true
142+ False
143
144Lesson 6E Introduction to Computer Vision: Post Quiz
145* Optical Flow helps us to understand how each pixel on video frames move.
146+ true
147- false
148* _____ computes the vector field that shows where each pixel is moving
149- Sparse Optical Flow
150+ Dense Optical Flow
151- none
152* Resizing and Blurring are steps that can be taken during?
153+ pre-processing
154- training
155- image transformation
156
157Lesson 7B Convolutional Neural Networks: Pre Quiz
158* To extract patterns from images we use?
159+ convolutional filters
160- extractor
161- filters
162* One of these is not a CNN Architecture
163- ResNet
164- MobileNet
165+ TensorFlow
166* CNN are mostly used for computer vision tasks.
167+ true
168- false
169
170Lesson 7E Convolutional Neural Networks: Post Quiz
171* Which pooling layer is used "scale down" the size of the image
172- average pooling
173- max pooling
174+ a and b
175* Convolutional networks generalizes much better
176+ True
177- False
178* To train our neural network, we need to convert images to tensors
179+ true
180- false
181
182Lesson 8B Pre-trained Networks and Transfer Learning: Pre Quiz
183* Transfer learning approach uses untrained models for classification
184- true
185+ false
186* One of these is not a normalization technique?
187+ height normalization
188- weight normalization
189- layer normalization
190* We choose Stochastic Gradient Descent(SGD) in deep learning because classical gradient descent can be ____
191- fast
192+ slow
193
194Lesson 8E Pre-trained Networks and Transfer Learning: Post Quiz
195* Dropout layers act as a ____ technique
196- gradient boosting
197- training
198+ regularization
199* freezing weights of convolutional feature extractor can be done by ____
200- setting `requires_grad` property to `False`
201- setting `trainable` property to `False`
202+ a and b
203* Batch normalization is to bring values that flow through the ____ to right interval
204- algorithms
205- batches
206+ neural network
207
208Lesson 9B Autoencoders: Pre Quiz
209* Self-supervised learning uses ____ data for training
210- pre-trained
211+ raw
212- labeled
213* Encoder Network coverts input images into latent spaces
214+ true
215- false
216* VAE is short for?
217- Variable AutoEncoding
218+ Variation auto-encoder
219- Variational automated encoders
220
221Lesson 9E Autoencoders: Post Quiz
222* Properties of autoencoders include
223- it is data Specific
224- works on unlabeled data
225+ all of the above
226* Auto encoders can be used to effectively remove noise from images
227+ true
228- false
229* Variational auto-encoders loss function does not consist of which of these?
230- reconstruction loss
231- KL loss
232+ TF loss
233
234Lesson 10B Generative Adversarial Networks: Pre Quiz
235* Generators take vectors and produce ____
236- videos
237+ image
238- gif
239* GANs is short for?
240- General adversarial networks
241- Generative advisor networks
242+ Generative adversarial networks
243* GAN uses ____ neural networks
244- 1
245+ 2
246- 3
247
248Lesson 10E Generative Adversarial Networks: Post Quiz
249* We can use Batch normalization and BatchNorm1D to stabilize the training
250+ true
251- false
252* Deep Convolutional GAN uses convolutional layers for ____ and ____
253+ generator, discriminator
254- CNN, generator
255- training, testing
256* Problems of GAN training includes
257- Sensitivity to hyperparameters
258- Keeping balance between generator and discriminator
259+ all of the above
260
261Lesson 11B Object Detection: Pre Quiz
262* Neural networks can only be used to classify images
263- true
264+ false
265* With object detection, we don't just get the class of an object, but also its ____
266- shape
267+ location
268- type
269* How many objects can an object detection model detect?
270- one
271- two
272+ any number
273
274Lesson 11E Object Detection: Post Quiz
275* An Object detection model gives us
276- object class
277- bounding box
278+ both class and bounding box
279* Which object detection models are faster?
280+ one-pass models
281- region proposal networks
282- Fast R-CNN
283* Which metric can be used to determine how well bounding boxes are aligned?
284- accuracy
285- precision
286+ IoU
287
288Lesson 12B Segmentation: Pre Quiz
289* How many segmentation algorithm are there?
290- 1
291+ 2
292- 3
293* Segmentation is a _____ task
294+ computer vision
295- natural language processing
296- neural networks
297* Segmentation networks consist of ____ and ____ parts
298- classifier, divider
299+ encoder, decoder
300- generator, discriminator
301
302Lesson 12E Segmentation: Post Quiz
303* ____ extracts features from an input image
304- decoder
305- generator
306+ encoder
307* ____ transforms input features into mask image
308+ decoder
309- generator
310- encoder
311* SegNet relies on ____ to train multi-layered network
312+ batch normalization
313- height normalization
314- weight normalization
315
316Lesson 13B Text Representation: Pre Quiz
317* Each word in a Bag of Words is linked to a vector index
318+ true
319- false
320* Text can be represented using _____ approaches
321- 1
322+ 2
323- 3
324* Character level representation represents each _____ as a number
325+ letter
326- word
327- symbol
328
329Lesson 13E Text Representation: Post Quiz
330* Word level representation represents _____ as a number
331- letter
332+ word
333- symbol
334* N-Grams refers to _____
335- combination of n number of words and symbols
336- combination of n number of letters
337+ combination of n number of Words
338* The main drawback of N-gram is that the vocabulary size grows fast
339+ true
340- false
341
342Lesson 14B Embeddings: Pre Quiz
343* Embedding is used to represent words with _____ dimensional dense vectors
344+ lower
345- higher
346- average
347* Word2Vec pre-trained embeddings can also be used in place of embedding layer in neural networks
348+ True
349- False
350* Using embedding layer we cannot switch from bag-of-words to embedding bag
351- True
352+ false
353
354Lesson 14E Embeddings: Post Quiz
355* Word2Vec has _____ main architectures
356- 1
357+ 2
358- 3
359* Word sense disambiguation is a limitation of traditional pretrained embedding representations
360+ True
361- False
362* An embedding layer takes _____ as input
363+ word
364- symbol
365- number
366
367Lesson 15B Language Modeling: Pre Quiz
368* Which of the following can be considered a language model?
369+ Word2Vec embeddings
370- Embedding layer in RNN
371- RNN used for text classification
372* A language model should be able to ____ the next word in the sentence
373- use
374+ predict
375- guess
376* Language models are trained on
377- language vocabulary
378- specially labeled data
379+ any natural text
380
381Lesson 15E Language Modeling: Post Quiz
382* Which of the architectures predicts a word from neighboring words?
383+ CBoW
384- Skip-gram
385- N-Gram
386* When we train a CBoW model, we obtain
387- A model that can generate text
388+ Word2Vec embedding vectors
389- Text classification model
390* The CBoW model is based on
391+ Dense neural network
392- Convolutional neural network
393- Recurrent neural network
394
395Lesson 16B RNN: Pre Quiz
396* RNN is short for?
397- regression neural network
398+ recurrent neural network
399- re-iterative neural network
400* A simple RNN cell has two weight _____
401+ matrices
402- cell
403- neuron
404* vanishing gradients is a problem of _____
405+ RNN
406- CNN
407- KNN
408
409Lesson 16E RNN: Post Quiz
410* _____ takes some information from the input and hidden vector, and inserts it into state
411- forget gate
412- output gate
413+ input gate
414* Bidirectional RNNs runs recurrent computation in _____
415+ both directions
416- north-west direction
417- left-right direction
418* All RNN Cells have the same shareable weights
419+ True
420- False
421
422Lesson 17B Generative networks: Pre Quiz
423* RNNs can be for generative tasks
424+ yes
425- no
426* _____ is a traditional neural network with one input and one output
427+ one-to-one
428- sequence-to-sequence
429- one-to-many
430* RNN generates texts by generating next output token for each input token
431+ true
432- false
433
434Lesson 17E Generative networks: Post Quiz
435* Output encoder converts hidden state into _____ output
436+ one-hot-encoded
437- sequence
438- number
439* Selecting the character with higher probabilities always gives a meaningful text.
440- true
441+ false
442- maybe
443* Many-to-many can also be referred to as _____
444- one-to-one
445+ sequence-to-sequence
446- one-to-many
447
448Lesson 18B Transformers: Pre Quiz
449* Attention mechanism provides a means of _____ the impact of an inout vector on an output prediction of RNN
450+ weighting
451- training
452- testing
453* BERT is an acronym for
454- Bidirectional Encoded Representations From Transformers
455+ Bidirectional Encoder Representations From Transformers
456- Bidirectional Encoder Representatives of Transformers
457* In positional encoding the relative position of the token is represented by number of steps
458+ true
459- false
460
461Lesson 18E Transformers: Post Quiz
462* Positional embedding _____ the original token and its position within the sequence
463- separates
464- compares
465+ embeds
466* Multi-Head Attention is used in transformers to give network the power to capture _____ of dependencies
467+ different types
468- same type
469- none
470* In transformers attention is used in _____ instances
471- 1
472+ 2
473- 3
474
475Lesson 19B Named Entity Recognition: Pre Quiz
476* What does NER stands for?
477- Nearest Estimated Region
478- Nearest Entity Region
479+ Named Entity Recognition
480* An entity always consists of one token
481- true
482+ false
483* To train NER model, we need
484+ Labeled dataset
485- Any natural text
486- Translated texts in two languages
487
488Lesson 19E Named Entity Recognition: Post Quiz
489* NER model is essentially a ____ model
490- text classification
491+ token classification
492- text regression
493* Which neural network types can be used for NER?
494- RNNs
495- Transformers
496+ Both RNNs and Transformers
497* NER model is a good example of ____ network architecture
498- one-to-one
499- one-to-many
500+ many-to-many
501
502Lesson 20B Language Models: Pre Quiz
503* What does GPT stand for?
504- Generic Pre-Trained network
505+ Generative Pre-trained Transformers
506- Generic Positional Text
507* What can GPT be used for?
508- Text generation
509- Text classification
510+ Both text generation and other tasks
511* GPT is based on transformer architecture
512+ true
513- false
514
515Lesson 20E Language Models: Post Quiz
516* What is zero-shot learning?
517+ Getting an answer from pre-trained network
518- Training the network from scratch
519- Training the network only for one epoch
520* Prompt engineering can be used with
521- Zero-shot learning
522- Few-shot learning
523+ Both
524* Which metric can be used to estimate the quality of a language model?
525- accuracy
526- recall
527+ perplexity
528
529Lesson 21B Genetic Algorithms: Pre Quiz
530* Genetic Algorithms are based on which of the following?
531- mutations
532- Selection
533+ both a and b
534* Crossover allows us to combine two solutions together to obtain a new valid solution
535+ true
536- false
537* Valid solutions to genetic algorithm can be represented as _____
538+ genes
539- neurons
540- cells
541
542Lesson 21E Genetic Algorithms: Post Quiz
543* Genetic Algorithms can solve which of these tasks
544- Schedule optimization
545- Optimal packing
546+ both of a and b
547* In implementing a genetic algorithm the first step is to randomly select two genes
548- true
549+ false
550* When using Crossover operation the algorithm randomly selects _____ genes
551- 3
552- 1
553+ 2
554
555Lesson 22B Reinforcement Learning: Pre Quiz
556* To train RL model, we need
557+ Simulation environment
558- Labeled dataset
559- Unlabeled dataset
560* What is a good example of Reinforcement learning:
561- Zero-shot image classification
562- Zero-shot text classification
563+ Learning to play chess
564* When creating RL-based chess engine, we need to
565- use all existing chess matches as a dataset
566+ let computer play against itself many times
567- program exhaustive search algorithm
568
569Lesson 22E Reinforcement Learning: Post Quiz
570* How does RL training algorithm knows how well it did?
571- It achieves high accuracy
572- Using perplexity metric
573+ Using reward function
574* Which problem(s) is RL is applicable to?
575- With discrete environment
576- With continuous environment
577+ Both
578* In Actor-Critic model, critic predicts
579+ Reward function
580- Best next action
581- Probability of next actions
582
583Lesson 23B Multi-Agent Modeling: Pre Quiz
584* By modeling the behavior of simple agents, we can understand more complex behaviors of a system.
585+ true
586- false
587* The principle of metasystem transition is derived from:
588- Evolutionary Cybernetics
589- Emergentism
590+ both of these
591* Multi-Agent systems emerged in the ____:
592- 1970s
593- 1980s
594+ 1990s
595
596Lesson 23E Multi-Agent Modeling: Post Quiz
597* An agent is:
598- an entity that lives alone
599+ an entity that lives in an environment
600- an entity that is intelligent
601* Reactive agents usually have:
602+ simple request-response behavior
603- complex behavior
604- no behavior
605* Multi-agent systems are used in:
606- video production and systems modeling
607- games and automations
608+ both the above
609
610Lesson 24B Ethical and Responsible AI: Pre Quiz
611* Why we need to worry about Ethical AI?
612- AI is a very powerful tool and can cause harm
613- We need to make sure AI models do not discriminate people
614+ Both
615* Which is the example of interpretable AI?
616+ Expert system
617- Neural network
618- Image classifier
619* It is not ethical to use AI in medicine
620- true
621+ false
622
623Lesson 24E Ethical and Responsible AI: Post Quiz
624* Why an AI model can discriminate?
625- Because it may become unfriendly
626+ Because datasets were not properly balanced
627- Because developers programmed it in such a way
628* Which of the following is not a principle of Responsible AI?
629- Transparency
630- Fairness
631+ Cleverness
632* Accountability of an AI system means that
633+ there should be a human being involved in taking decisions, who can take responsibility
634- AI system should be held responsible for its actions
635- AI system developers should be held responsible
636* Model fairness is related to
637- Interpretability
638+ Biases
639- Accountability