Exam questions

Question 1

Definition of CS

Answer 1

Autonomous system that can perceive its environment, learn from experience, anticipate the outcome of events, act to pursue goals and adapt to changing circumstances.

Question 2

Purpose of Cognition

Answer 2

Cognition enables a system to operate in a meaningful way beyond its original pre-programmed behavior and specification and to deal with uncertainty.

Question 3

Difference between cogition and intelligence

Answer 3

Cognition is a global process, integrating different process modalities. Intelligence (and learning, memory, interaction) are cognitive skills

Question 4

Paradigms of cognitive modeling

Answer 4

Cognitivism: Hypothesis that cognition is a form of computation. Cognitive functions are modelled as working computer programs
Emergentism: Cognition is a continuous self-organizing process, driven by interaction between the agent and its environment.

Question 5

Cognitive capabilities

Answer 5

Self-reliance:

• autonomy
• interaction
• goal-directedness

Perception and action:

• interpretation
• sensing
• anticipating
• action

Adaptation:

• reaction
• learning
• anomaly detection

Question 6

Difference between bio-inspired model and purely computational model

Answer 6

Computational Model: implements cognitive functions soley based on a functional view of the system without any reference to biology
Bio-inspired model: implement cognitive functions by replicating known or hypothesized mechanisms of cognitive processing in biological organisms.

Question 7

Explain Robotics Paradox and two different example environments

Answer 7

Robots excel at well-defined repeatable tasks where they can outperform even the most skilled human workers

No robot built up to now can operate in dynamic real-world environments and carry out simple everyday tasks, such as walking, shopping etc.

Complexity of simple tasks: technical systems operate to exact formal descriptions –> no formal description of simple tasks such as shopping, because there is no formal description of the world.

Examples:
Welding in a highly controlled environment like a car factory: Well defined, robots can outperform most skilled human.

Shopping in a dynamic environment like a crowded mall: Going to shop is different every time (shopping list, place of items, traffic)

Question 8

Purpose, process and succeed-failure evaluation of Turing Test

Answer 8

Purpose: used to determine if computer can think
Process: Interrogator communicates via notes with a person and machine. If interrogator cannot tell which one of the two is the computer, Turing Test is passed.
Success-failure evaluation: If the Computer manages to deceive the interrogator as often as the Human, it succeeds, else it fails.

Question 9

Ultimate and Proximate Distinction. Give the 2 Questions

Answer 9

ultimate distinction: Why does a cognitive system exhibit a certain behavior?

Crying elicits care and defense from caregivers. Babies that do not cry when in need have a lower chance of surviving

proximate distinction: How is a certain behavior implemented in a cognitive system?

Babies cry because of separation from their caregivers, cold, lack of food, and other internal mechanisms

Question 10

Give (a) structural and function brain image method in full words

Answer 10

structural:
Magnetic Resonance Imaging (MRI)
Diffusion Tensor Imaging (DTI)

functional:
Electroencephalography (EEG)
Magnetoencephalography (MEG)
Positron Emission Tomography (PET)
Single-Photon Emission Computer Tomography (SPECT)
functional Magnetic Resonance Imaging (fMRI)

Question 11

Cycle of cognitive processing

Answer 11

Cognition (Anticipation -> assimilation -> adaptation) -> action -> perception - > cognition

Question 12

Degrees of biological realism?

Answer 12

Computational Models
- Purely computational models implement cognitive functions solely based on a functional view of the system without any reference to biology.
Bio-Inspired Models
- Bio-inspired models implement cognitive functions by replicating known or hypothesized mechanisms of cognitive processing from biological organisms.
Hybrid Models
- Hybrid models combine computational and bio-inspired modeling.

Question 13

Common name of the different anatomical brain regions

Answer 13

• gray matter
• white matter

(Frontal lobe, parietal lobe, occipital lobe, temporal lobe)

Question 14

Which anatomical structure plays the key role in the formation of episodic long-term memories?

Answer 14

Hippocampus

Question 15

Name the three? ( should be 4) lobes for each lobe, name (one) cognitive function that it implements

Answer 15

• frontal lobe: short-term memory
• parietal lobe: somatic sensation
• temporal lobe: hearing
• occipital lobe: vision

Question 16

Difference sympathetic and parasympathetic nervous system

Answer 16

sympathetic: prepare the organism for stress, connect internal organs to the brain
parasympathetic: set the body to resting state and increase digestive functions

Question 17

Difference between static and dynamic connectivity in the brain

Answer 17

static (anatomical connectivity) and

dynamic (functional connectivity, dependent on the current cognitive task)

Question 18

Is it true that only 10% of the brain is used?

Answer 18

Every part of the brain is used but not at once otherwise epilepsy would occur.

Question 19

What function does the Thalamus have

Answer 19

Relay and distribution of sensory and motor signals to the different regions of the cerebral cortex.

Question 20

What function does the Epithalamus have?

Answer 20

production melatonin

Question 21

What function does the Hypothalamus have?

Answer 21

Control of autonomic functions (temperature regulation, appetite), behavior, and
hormone production

Question 22

What is the other nervous system next to the central nervous system and its subsystems and their purpose?

Answer 22

Peripheral Nervous system:
The autonomic nervous system is responsible for homeostasis (self-regulation) and operates largely unconsciously. It includes two antagonist subsystems:

Sympathetic nervous system:
- connection of internal organs to the brain; preparation of the organism for stress
Parasympathetic nervous system:
- mainly cranial nerves and lumbar spinal nerves; sets body to resting state and increases digestive functions.

Question 23

4 differences between spiking neuron models and normal neuron models

Answer 23

model: dynamical system - activation function
communication: digital spikes (all or none) - analog values (real numbers)
encoding: spike time, rate, synchrony - real numbers
computational complexity: high -low
biological plausibility: usually high - usually low

Question 24

four phases of a neuron

Answer 24

resting state
depolarization
repolarization
refactory period

Question 25

For a standard leaky integrate-and-fire neuron, write a dynamics function of the membrane potential for resting potential.

Answer 25

Write down equation from studocu

Question 26

Name learning rules and how they are realized in biological neurons. Opposite learning rules with opposite effects.

Answer 26

Hebbian learning: learning rule that reinforces causal relations between events

When an axon of cell A is near
enough to excite cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A’s efficiency, as one of the cells firing B, is increased.

Question 27

What is signal encoding and what is signal filtering?

Answer 27

Signal encoding: learn to produce a desired output spike train
Signal filtering: extract relevant features and discard undesired information.

Question 28

3 Paradigms of ML

Answer 28

Supervised learning
Unsupervised learning
Semi-supervised learning

Question 29

How to deal with overfitting?

Answer 29

• Regularization
• More data (with noise)
• data augmentation

Question 30

Explain cross-validation

Answer 30

• fold data into k subsets and the data is learned in k iterations
• in every iteration a different subset is selected as the validation set
• overall performance corresponds to the averaged performance of k iterations

Question 31

Advantages of SVM over Least Squares regarding separation of clusters

Answer 31

• SVM is not sensitive to outliers

- SVM minimizes the generalization error by computing the hyperplane that maximizes the classifier margin.

Question 32

Interpolation vs regression

Answer 32

Interpolation:

• given n data points, when you interpolate, you look for a function that is of some predefined form that has the values in that points exactly as specified.
• when you do regression, you look for a function that minimizes some cost, usually sum of squares of errors. You don’t require the function to have the exact values at the given points, you just want a good approximation.

Question 33

Analog and spiking neuron models

Answer 33

Analog neuron models are functions that are not able to produce the temporal dynamics of biological neurons.
Spiking neural networks (SNNs) are artificial neural networks that are more closely mimicing natural neural networks. In addition to neuronal and synaptic state, SNNs incorporate the concept of time into their operating model. Neurons in SNN do not fire at each propagation cycle, but rather fire only when a membrane potential reaches a specific value.

Question 34

Neurorobotics

Answer 34

Is the science and technology of embodied autonomous neural systems. A neural system is any type of simulated brain model or even an actual biological brain, the embodiment is either a physical robotic system or any simulated virtual body.

Question 35

Two theoretical implementations for embodiment

Answer 35

• simulated virtual embodiment

- physical robotic system

Question 36

Give 5 classic human modalities and one additional sense of an animal

Answer 36

• light sense: vision
• mechanical sense: touch, hearing
• chemical sense: taste, smell
• animal sense: magnetic sense

Question 37

Name and explain two low-level motor control functions that are implemented by the spinal cord

Answer 37

Spinal Reflexes:
- coordinate motor responses to a stimulus applied to peripheral receptors
Central Pattern Generation:
- are neuronal spinal networks capable of generating rhythmic motor activity without input from peripheral receptors

Question 38

Saccades

Answer 38

Saccades refer to both voluntary and reflexive eye movements that direct the gaze towards objects and regions of interest.

Question 39

How does the Turing test deal with the meaning of symbols? (Answered by me)

Answer 39

Even if the Turing Machine is able to decept the interrogator a large amount of times it does not have to mean that the machine actually understands the meaning of symbols. Like in the Chinese room example it can be that the Turing machine operates on symbols with rules but does not understand them.

Question 40

Uncanny valley

Answer 40

a steady increase in “human- likeness” of a robot does not yield a steady increase of familiarity to humans

Question 41

What makes real world planning difficult? 2 arguments

Answer 41

• ever changing environment

- no formal description of simple tasks in the real world. (location of objects change etc.)

Question 42

Is it advantageous for a cognitive system to have a body?

Answer 42

Embodied cognition is an approach to cognition that has roots in motor behavior. This approach emphasizes that cognition typically involves acting with a physical body on an environment in which that body is immersed.

Question 43

How is information stored in spike trains?

Answer 43

Rate Coding: The information is carried in the averaged spike rate within a certain time window (e.g. muscle excitation)
Time-to-First Spike Coding:
The information is encoded in the time of emission of the first spike after a reference point (e.g. onset of a stimulus).

Question 44

explicit learning

Answer 44

Explicit learning is equivalent with intentional learning of information

Question 45

What kind of learning uses labeled data?

Answer 45

Supervised learning

Question 46

2 advantages of simulating robots?

Answer 46

• increased simulation speed

- no need for expensive robots