stud
bud
Sign up
Log in
Home
Library
Explore
SAT
More
Final IBCS Quiz - paper 1 - 3 Flashcards | StudBud
Explore
/
Final IBCS Quiz - paper 1 - 3
Final IBCS Quiz - paper 1 - 3
200 terms
·
StudBud Administration
·
1 play
Imported from external platform
Flashcards
200 cards
F
Practice quiz
P
Host live
L
Questions
200
Answers
1
What does multi-head attention achieve compared to single-head attention?
Increasing model parameters
Slower inference
Simplified model training
Richer representation of input relationships
2
In RNNs, what does each time step represent?
A random data point
An output layer
A single sequence element
A model checkpoint
3
Why are recurrent loops not needed in Transformer architectures?
Transformers are slower than RNNs
Transformers avoid sequential dependencies
Transformers use larger datasets
Transformers simplify gradient descent
4
In terms of LLMs, what is one consequence of training on low-quality data?
Smaller model size
Hallucination and inaccurate outputs
Improved generalization
Faster convergence
5
What design flaw in LLMs can cause 'hallucination' problems?
Lack of grounding in factual data
Poor gradient clipping
Excessive parameter tuning
Too much dropout
6
Which type of logic gate outputs 1 only if inputs are different?
NOT
XOR
OR
AND
7
What type of testing involves executing the code?
White Box Testing
Static Testing
Dynamic Testing
Beta Testing
8
What information does positional encoding provide in a Transformer?
Word meaning
Word frequency
Word position
Word importance
9
What is Bag-of-Words?
Representing text by word frequencies
Image recognition
Grammar-based translation
Sorting sentences alphabetically
10
Which pseudocode best describes a linear search algorithm?
Use two pointers to scan from both ends
Binary search using midpoints
Sort the array before searching
Loop through each element until a match is found
11
What does 'tokenization' refer to in NLP?
Combining sentences
Encrypting text
Splitting text into words or pieces
Reordering words
12
Which memory type holds instructions needed to boot up a computer?
ROM
Hard Drive
RAM
Cache
13
Why are GPUs better than CPUs for deep learning?
They have higher memory
They are cheaper
They can parallelize computations
They are easier to program
14
Which of the following devices would most likely have a NIC installed?
Smartphone
USB Drive
Printer
Laptop
15
Which layer in a neural network gives final prediction output?
Hidden layer
Dropout layer
Input layer
Output layer
16
What is the main advantage of a TPU over a CPU for training LLMs?
Higher memory capacity
Smaller size
Higher matrix computation speed
Lower cost
17
Which of the following is an example of secondary memory?
Cache
RAM
Hard Drive
ROM
18
In a binary search algorithm, if the array is unsorted, the search will:
Run faster
Always succeed
Automatically sort the array first
Fail or give incorrect results
19
Which backup strategy ensures a minimal data loss during a sudden failure?
Weekly backup
Real-time backup
Monthly backup
No backup
20
In a critical system, why must model training and inference data be isolated?
To prevent data leakage and biased evaluation
To improve learning rate
To accelerate hardware parallelism
To increase training batch size
21
In RNNs, why does gradient explosion occur?
Learning rate is too small
Too few hidden layers
Gradients are multiplied by large values over time steps
Cell states are ignored
22
Which type of bias occurs when outdated information affects training?
Labeling bias
Historical bias
Sampling bias
Confirmation bias
23
What can happen if transformer models lack sufficient training data diversity?
Higher generalization ability
Biased or narrow outputs
Reduced model size
Faster training time
24
What is the primary purpose of cache memory?
To run the operating system
To provide a backup storage
To permanently store files
To temporarily store frequently used data
25
What is the major advantage of a doubly linked list over a singly linked list?
Lower memory usage
Traversal in both directions
Requires fewer pointers
Faster insertion at the end only
26
In pseudocode, what is the purpose of the resetNext() method for a collection?
To clear the collection
To delete the first element
To find the maximum value
To move pointer to the start
27
Why might TPUs outperform GPUs in training large LLMs?
TPUs are cheaper
TPUs have smaller memory
TPUs have fewer cores
TPUs are optimized for matrix operations
28
Which situation best describes the use of a closed-loop control system?
A refrigerator adjusting temperature based on internal sensors
A printer printing pages continuously
A microwave operating for a preset time without feedback
A coffee machine dispensing coffee once a button is pressed
29
Why are fiber optic cables preferred over copper for internet backbones?
Fiber has higher speed and longer transmission distance
Fiber is easier to install
Fiber is cheaper than copper
Fiber is more susceptible to electromagnetic interference
30
Why is cross-validation important in ML?
It helps evaluate generalization
It tests model on training data
It reduces dataset size
It speeds up inference
31
Which problem may occur if virtual memory is heavily used?
Disk thrashing
Better cache performance
Faster processing
Increase in RAM capacity
32
Which type of transmission medium is wireless?
Coaxial Cable
Ethernet Cable
Radio Waves
Fiber Optic
33
What does SaaS stand for?
Software as a Service
Service as a Software
Software and Advanced Security
Storage and Application Software
34
What is the function of an actuator in a control system?
To detect environmental changes
To convert electrical signals into physical motion
To store sensor data
To process digital signals
35
In Transformers, what does multi-head attention allow the model to do?
Reduce training epochs
Process sentences faster
Generate synthetic training data
Attend to different representations simultaneously
36
What type of loop repeats as long as a condition remains true?
While loop
Do-then loop
For loop
Infinite loop
37
Why is explainability critical in AI decision systems?
To improve training speed
To reduce GPU memory needs
To increase bias in outputs
To satisfy user trust and accountability
38
What happens first when an interrupt is received by the CPU?
The CPU saves the current state
The program continues to run normally
The CPU immediately executes the next process
The CPU increases its clock speed
39
In the context of data migration, what is a major risk when moving between systems?
Increase in processor speed
Decrease in network bandwidth
Enhanced database integrity
Incompatible file formats
40
Which device is used to amplify and boost Wi-Fi signals?
Hub
Router
Repeater
Switch
41
Which action occurs during dynamic testing but not during static testing?
Executing program code
Reviewing documentation
Validating project requirements
Verifying user manuals
42
In a control system, which sensor would most likely be used to detect water levels?
Ultrasonic sensor
Tilt sensor
pH sensor
Infrared sensor
43
Which of the following is the correct order for a typical IPO (Input-Process-Output) model in a control system?
Process → Input → Output
Input → Process → Output
Process → Output → Input
Output → Input → Process
44
Which network type typically covers a single building?
LAN
VLAN
WAN
SAN
45
What effect does using high-frequency tokens in a dataset have on a model?
Improves diversity
Skews learning towards common patterns
Reduces bias
Enhances rare event detection
46
Which of the following is an example of an actuator?
LED light
Mouse
Monitor
Keyboard
47
Which logic gate outputs 1 only if both inputs are 1?
NOT
OR
XOR
AND
48
Which of the following best defines a firewall?
A device that boosts Wi-Fi
A tool for encrypting hard drives
A physical wall that blocks signals
A software/hardware that controls network traffic
49
Which is an NLP example?
Spotify
Face ID
Photoshop
Google Translate
50
What is the main challenge when applying dropout to attention layers?
Reduces input size
Causes loss of critical relational patterns
Forces higher learning rate
Simplifies model architecture
51
What is the primary purpose of backpropagation?
Increase model loss
Adjust weights to minimize loss
Speed up predictions
Store previous outputs
52
Which of the following is an example of an operating system?
Google Chrome
Microsoft Word
Windows 10
Adobe Photoshop
53
What does 'attention weights' represent in a self-attention mechanism?
Number of neurons activated
Batch size value
Relationship strength between input tokens
Size of the dataset
54
What challenge can arise from having too many layers in a neural network?
Underfitting
Overloading GPU memory
Vanishing or exploding gradients
Faster convergence
55
What does 'synthetic data' mean?
Data stored in compressed files
Data collected from sensors
Data created from real surveys
Data artificially generated
56
Why is the critical path important when optimizing a ML pipeline?
It shows model accuracy
It estimates model loss
It predicts training data size
It identifies the bottleneck tasks
57
What is the full form of CPU?
Computer Processing Unit
Central Processing Unit
Computer Programming Unit
Central Programming Unit
58
Which protocol is used for secure web browsing?
FTP
HTTP
HTTPS
TCP
59
What is the output of the following pseudocode? X=3 Y=5 output X+Y
2
15
35
8
60
What is the first step in the Fetch-Decode-Execute cycle?
Decode
Execute
Store
Fetch
61
Which of the following represents a benefit of automated testing?
Testing is less reliable
Testing is slower
Testing requires manual verification
Testing can be repeated consistently and quickly
62
What aspect of a critical path algorithm helps reduce machine learning deployment time?
Increase memory speed
Prioritize longest dependent tasks
Reduce model size
Increase hidden layers
63
In what case would a chatbot be considered ethically dangerous?
When it answers too quickly
When it manipulates emotional users
When it uses formal language
When it only gives short responses
64
Which part of the CPU stores the address of the next instruction?
PC
ALU
MAR
MDR
65
Which is an advantage of SaaS (Software as a Service)?
Users control hardware
Access from anywhere with internet
Users control updates
No internet needed
66
What does 'Bag-of-Words' fail to capture?
Word count
Vocabulary size
Word order and context
Word frequency
67
Which pseudocode structure would best find the maximum value in an array?
Infinite while loop
While loop without comparison
For loop with no conditions
For loop with conditional checking
68
What does the input gate in an LSTM control?
Forgetting old memory
Outputting predictions
Adding new information
Incoming hidden states
69
Which data structure is most suitable for undo functionality in applications?
Linked List
Stack
Graph
Queue
70
Which memory is directly accessed by the CPU for fast data retrieval?
ROM
Cache
RAM
Hard Drive
71
What can cause training slowdown in a GPU cluster?
Too many hidden layers
Poor network connection between GPUs
Use of smaller datasets
Lower batch sizes
72
In the context of the IPO model, feedback is used to:
Repeat the same output
Adjust inputs based on output
Bypass processing step
Increase output speed
73
Which testing method checks internal paths and logic of the code?
Black Box Testing
Beta Testing
Alpha Testing
White Box Testing
74
In terms of memory, what is a major advantage of TPUs over GPUs?
TPUs can store entire datasets
TPUs have smaller bandwidth
TPUs have much larger VRAM
TPUs have optimized memory for matrix ops
75
What problem does the self-attention mechanism in Transformers primarily solve?
Memory optimization
Capturing long-range dependencies
Increasing network depth
Parallel computation
76
Which data structure is best for implementing a breadth-first search (BFS)?
Tree
Linked list
Stack
Queue
77
Why do deep networks require careful weight initialization?
To avoid vanishing or exploding gradients
To reduce number of parameters
To accelerate overfitting
To simplify loss calculation
78
What happens if GPU memory is full during model training?
Training continues normally
More batches are processed at once
Training crashes or slows down
Training speeds up
79
Which bias results from non-representative training samples?
Labeling bias
Sampling bias
Historical bias
Confirmation bias
80
What problem can occur if training data overlaps with testing data?
Overfitting
Lower training speed
Increased validation accuracy
Underfitting
81
In a distributed system, which of the following is TRUE?
A single point of failure affects the whole system
Workload is shared across multiple nodes
It requires no network connectivity
Only one processing unit controls all operations
82
What problem does BPTT cause in RNNs?
Low latency
Memory loss
Vanishing gradient
Overfitting
83
Which term refers to the smallest unit of data in computing?
Nibble
Bit
Byte
Word
84
What does a loss function measure?
Data size
Difference between prediction and actual output
Network speed
Memory usage
85
Which resource management issue occurs when RAM is full?
Immediate shutdown
Erasure of hard drive data
Increased CPU clock speed
Paging to secondary storage
86
Why are vanishing gradients less problematic in Transformers compared to RNNs?
Slower forward pass
No use of activation functions
Use of residual connections and parallelism
Fewer parameters
87
Why is vanishing gradient more severe in deep RNNs than in feedforward networks?
Smaller datasets
Lack of input features
Larger learning rate
Multiplication over time steps and layers
88
What is the hexadecimal value of decimal 15?
D
F
E
G
89
What could happen if a chatbot is trained without ethical oversight?
Improved efficiency
Bias amplification and misinformation spread
Reduced inference time
Faster training cycles
90
What occurs if a race condition happens in a multithreaded program?
Improved memory efficiency
Decreased CPU usage
Unpredictable behavior
Deterministic output
91
What kind of software is Microsoft Word?
Operating System
Device Driver
Application Software
Utility Software
92
Which keyword is typically used to create a loop in pseudocode?
input
loop
if
output
93
What does gradient descent do?
Store input data
Find optimal weights
Increase model errors
Calculate sentence length
94
What type of data does a Recurrent Neural Network process best?
Sequential text
Static images
Sorted tables
Random numbers
95
In a chatbot scenario, what ethical issue is linked to leaking personal data from training sets?
Latency optimization
Resource scaling
Bias mitigation
Latency optimization
96
Which operator checks equality in pseudocode?
=
=>
==
!=
97
What does RAM stand for?
Real Allocation Management
Read Access Memory
Random Access Memory
Rapid Action Machine
98
Which validation method checks if two fields match?
Consistency check
Length check
Presence check
Format check
99
Why is it dangerous to overfit a language model on synthetic data only?
Model needs more hidden layers
Model needs fewer neurons
Model memorizes unrealistic patterns
Model becomes faster but less accurate
100
Which aspect of dataset preparation can help minimize selection bias?
Only using synthetic data
Random sampling across all user groups
Filtering out incorrect labels
Increasing model parameters
101
How can historical bias affect an insurance chatbot’s performance?
It reduces model size
It improves fairness
It speeds up model training
It causes errors based on outdated practices
102
What happens if a base case is missing in a recursive function?
The function runs faster
The function loops infinitely
The function immediately stops
The function sorts the input
103
Which of the following is a characteristic of a doubly linked list?
Each node points to both the next and previous nodes
The list has no head or tail
Each node points only to the previous node
Each node points only to the next node
104
In pseudocode, which command is used to receive input from a user?
input
scan
print
output
105
Which of the following best defines a stakeholder?
A user who owns the server
A device that connects a network
A person with an interest in a project
A backup server administrator
106
If the network latency is very high, which of the following symptoms would appear first?
Delayed response times
High download speed
Stable network connections
Increased bandwidth
107
What is machine learning?
Using robots for manual work
Training a machine manually
Programming computers manually
Teaching machines to learn from data
108
Why is pragmatic analysis important for customer-facing chatbots?
It optimizes latency
It interprets meaning beyond literal words
It reduces training loss
It compresses input sequences
109
Which component stores long-term memory in LSTM cells?
Cell state
Forget gate
Output gate
Hidden state
110
How can hardware bottlenecks limit the performance of a GPU cluster?
Insufficient network bandwidth between GPUs
Faster CPU speeds
Larger batch sizes
More model layers
111
Which component carries out arithmetic and logical operations?
RAM
ALU
ROM
Cache
112
What is the role of the forget gate in an LSTM cell?
Store weights
Add information
Remove unnecessary information
Create output
113
Which feature of a high-level programming language improves program portability across different systems?
Human-readable Syntax
Access to machine instructions
Strong Typing
Hardware Abstraction
114
Which memory is the fastest but most expensive?
Cache
Hard Drive
Virtual Memory
RAM
115
What does a router mainly use to forward data packets to the correct destination?
IP Address
Serial Number
MAC Address
Device Name
116
What technique ensures that data sent over an unreliable network still arrives correctly?
VPN tunneling
Error detection and correction
Encryption
Packet loss tolerance
117
What improves chatbot responses in emotionally sensitive situations?
Training only on synthetic data
Ignoring emotional cues
Using emotional human conversations as training data
Using random datasets
118
How does a Transformer model process input sentences?
Word-by-word sequentially
Random words
Entire sentence at once
Layer-by-layer only
119
Which statement about recursion is TRUE?
Recursive solutions are always more efficient than iterative ones
Recursive functions must have a base case
Recursion does not require memory overhead
Recursive functions cannot be translated into loops
120
What do hidden layers do?
Compress input data
Analyze and transform data
Save training history
Display output
121
Which of the following is a non-volatile memory?
Cache
RAM
Virtual Memory
ROM
122
Which of the following is NOT an input device?
Monitor
Mouse
Keyboard
Scanner
123
In a control system, what is the role of a sensor?
To display information
To store data
To detect physical changes
To process signals
124
What risk arises when emotional language is not properly handled in chatbot responses?
Faster query processing
Miscommunication or customer distress
Reduced computational cost
Improved model validation
125
Which device connects multiple networks together?
Modem
Router
Hub
Switch
126
When using packet switching, which factor is critical for reassembling packets correctly?
Router speed
Packet size
Packet sequence numbers
Packet encryption
127
Which hardware is specially optimized for matrix calculations?
GPU
SSD
TPU
CPU
128
What ethical principle requires AI systems to avoid amplifying discrimination?
Bias mitigation
Latency optimization
Data privacy
Resource scaling
129
In backpropagation through time (BPTT), how does gradient shrinkage happen?
Skipping loss calculation
Using large batch sizes
Multiplying small derivatives across many time steps
Using large learning rates
130
Why are positional encodings needed in Transformer models?
To increase model size
To store training data
To speed up GPU processing
To preserve word order
131
Which factor increases inference latency?
More model layers
Faster network speeds
Smaller datasets
Fewer parameters
132
Which best describes the time complexity of binary search in a sorted array?
O(n log n)
O(1)
O(log n)
O(n)
133
What is the effect of the vanishing gradient on model training?
Faster convergence
No effect
Slower or failed learning
Better accuracy
134
Which type of bias can result from selecting data that fits preconceived ideas?
Sampling bias
Historical bias
Labeling bias
Confirmation bias
135
What is a critical path algorithm used for?
Minimize project completion time
Maximize training loss
Reduce data loss
Improve storage space
136
What is the vanishing gradient problem?
Gradients become too small during training
Network memory overflow
Too fast model convergence
Data loss in storage
137
What is an epoch in training?
The final test
One training cycle through all data
Loss calculation method
A single data point
138
Why is cooling important for GPU clusters?
Improve network speed
Enhance model accuracy
Reduce storage cost
Prevent performance throttling
139
In pseudocode, which structure is used for making decisions?
array
while
loop
if-then-else
140
How does batch size influence the stability of gradient descent?
Smaller batch size causes noisier updates
Batch size has no impact
Only affects memory usage
Larger batch size causes noisier updates
141
How does dropout regularization help prevent overfitting?
By randomly removing neurons during training
By increasing model complexity
By increasing weight values
By expanding the training dataset
142
In pseudocode, what happens if the condition in a while loop is never false?
The loop never runs
The program immediately crashes
The loop runs indefinitely
The loop runs once
143
What specific role do residual connections play in Transformer architectures?
Decrease inference speed
Bypass layers to preserve gradients
Increase model size
Enhance vanishing gradients
144
Which of the following best defines packet loss?
A packet being dropped during transmission
A packet arriving earlier than expected
A packet containing no data
A packet being duplicated
145
Which field involves machines understanding human language?
Data Mining
Computer Vision
Natural Language Processing
Robotics
146
What is the main purpose of an IP address?
To encrypt data
To uniquely identify devices on a network
To physically connect devices
To store user passwords
147
Which sorting algorithm works by repeatedly finding the minimum element and placing it at the beginning?
Insertion Sort
Merge Sort
Selection Sort
Bubble Sort
148
Which device is essential for connecting two different networks together?
Access Point
Hub
Switch
Router
149
How can transformer models handle very long input sequences efficiently?
Increase number of neurons
Apply hierarchical attention
Use recurrent memory units
Compress sequences using positional encoding
150
What is a drawback of using extremely large LLMs in production systems?
Higher inference latency and resource costs
Lower validation loss
Lower overfitting risk
Simpler model deployment
151
How does a larger batch size affect training?
Reduces memory usage
Improves long-term dependency learning
Speeds up convergence
Increases VRAM requirement
152
What does this pseudocode do?
4
1
5
2
153
In a peer-to-peer (P2P) network, each device acts as:
Both a server and a client
Only a client
Neither server nor client
Only a server
154
Which gate in LSTM decides what information to discard?
Memory gate
Forget gate
Output gate
Input gate
155
Which factor is most critical for avoiding bias when building training datasets?
Faster hardware
Larger dataset size
Higher model complexity
Representative sampling across groups
156
What is the primary weakness of Bag-of-Words representation?
Generates biased predictions
Ignores grammar and context
Requires too much training time
Needs large GPU memory
157
What is the decimal equivalent of binary 1010?
8
10
11
9
158
Which of the following best describes a network switch?
It connects multiple networks together
It routes packets to correct IP addresses
It broadcasts data to all connected devices
It connects devices in a network and sends data only to intended devices
159
Which traversal method processes the root node first before its subtrees?
Preorder
Postorder
Breadth-First
Inorder
160
In machine learning, what is a hyperparameter?
A network output
A fixed input
A training example
A tunable training setting
161
What does a self-attention mechanism do?
Sorts words alphabetically
Ignores unimportant words
Calculates relations between words
Removes punctuation
162
How does a too-small learning rate affect neural network training?
Reduces generalization
Causes very slow convergence
Speeds up convergence
Increases training loss permanently
163
Which network handles sequential data?
GAN
CNN
RNN
DNN
164
In a stack, which element is removed first?
Bottom
Random
Middle
Top
165
Which of the following is a major reason for using modular programming?
To duplicate code
To make programs longer
To improve maintainability and readability
To hide code functionality
166
How do transformer models maintain word order information?
Using recurrent loops
Hidden layers automatically learn order
Using positional encodings
Using sorted input vectors
167
What happens if dropout is applied too aggressively?
Model accelerates learning
Only output layer is affected
Model overfits faster
Model underfits and struggles to learn patterns
168
What happens if the learning rate is too high?
No change in model behavior
Slower training time
Model converges too early to suboptimal solution
Faster convergence to perfect model
169
Which of the following best defines a VLAN?
A wireless network inside a building
A network that covers multiple countries
A network that connects devices over Bluetooth
A logical segmentation of a LAN into isolated parts
170
What happens during the 'Decode' step in the Fetch-Decode-Execute cycle?
The instruction is executed
The instruction is interpreted
The instruction is fetched from memory
The program counter is incremented
171
Which of the following is a key reason for using a linked list instead of an array?
Arrays use less memory
Linked lists allow dynamic memory allocation
Arrays are more flexible than linked lists
Linked lists are faster for random access
172
What advantage does a Transformer have over an RNN?
Faster training via parallelization
Lower computational needs
Higher storage capacity
Simpler model architecture
173
What would be an ethical risk if chatbots predict user emotions incorrectly?
Model compression failure
Reduced hardware costs
Faster model convergence
Privacy breaches and emotional harm
174
What is one cause of vanishing gradients?
Multiplying small derivatives across layers
Increasing model accuracy
Adding dropout layers
Using large batch sizes
175
In networking, what does encryption mainly protect?
Data storage
Data structure
Data backup
Data confidentiality
176
Which storage device is typically used for long-term data storage?
ROM
Cache
Hard Drive
CPU
177
Which factor mainly impacts the effectiveness of a cache memory?
Cache location
Cache operating system
Cache encryption
Cache hit rate
178
In a stack, which operation adds an item?
Dequeue
Pop
Enqueue
Push
179
What helps an LSTM network retain information across long sequences?
Cell state
Activation function
Bag-of-words algorithm
Dropout layer
180
What does gradient descent minimize during training?
Processing time
Loss function
Model complexity
Data size
181
What does the acronym LAN stand for?
Logical Area Node
Large Access Node
Local Access Network
Local Area Network
182
Which of the following is a type of primary memory?
RAM
Optical Disk
Hard Drive
SSD
183
Which of the following best describes packet switching?
Data divided into packets and sent independently
Data broadcasted to all devices
Data sent through only one fixed path
Data sent as a single large file
184
Which of the following methods most directly prevents data loss due to hardware failure?
Cloud computing
Virtual memory
Increased CPU speed
Redundant systems
185
Why is gradient clipping sometimes used in RNN training?
To force early convergence
To increase inference latency
To prevent gradient explosion
To amplify small gradients
186
In which layer of the OSI model does data encryption typically occur?
Data Link
Presentation
Physical
Transport
187
Which of the following is a purpose of validation rules in databases?
To ensure data integrity
To display output
To create new records
To delete data
188
Why might relying solely on synthetic data introduce bias?
It accelerates model convergence
It captures too much variation
It prevents overfitting
It underrepresents real-world complexity
189
In the OSI model, which layer is responsible for logical addressing like IP addresses?
Data Link
Transport
Session
Network
190
What mechanism allows Transformers to capture word relationships?
Loss minimization
Gradient descent
Attention weights
Hidden states
191
Why are attention mechanisms superior to simple recurrence in modeling sequences?
They reduce training dataset size
They capture long-range dependencies directly
They limit inference speed
They increase memory usage
192
Which of the following best describes a collection in pseudocode?
A set of unique elements
A single character
A memory address
A single integer
193
In binary, what is the result of 1 + 1?
10
1
11
0
194
What does GUI stand for?
Graphical User Interface
General User Interface
Graphic User Information
Global User Interaction
195
Which of the following is a disadvantage of a centralized control system?
Easier to maintain
Lower cost
Higher flexibility
Failure of main controller leads to system failure
196
What makes TNNs (Transformer Neural Networks) more scalable than RNNs?
Lower model size
Parallelizable attention mechanisms
Sequential computation
Fixed-length memory
197
What is the main goal of using a validation set?
Save computation time
Test final performance
Tune hyperparameters
Train the model
198
How do LSTM gates collectively combat the vanishing gradient problem?
Increase data shuffling
Reduce batch size during training
Maintain and selectively update cell states
Regularize model outputs
199
What type of compression would be preferred for compressing a text document?
Lossless
Hybrid
Lossy
Compressed Video
200
Why can synthetic data sometimes harm model generalization?
Slower model convergence
Too complex patterns
Excessive training samples
Lack of real-world noise and variation
