Quiz
Ohne Titel
Erstelle ein kostenloses Konto, um dieses Quiz zu beenden
Speichere deinen Fortschritt und sieh dir dein Endergebnis an.
oder mit deiner E-Mail registrieren
Schon ein Konto? Anmelden
Fragen in diesem Quiz (44)
1Reinforcement Learning
WHAT IS?- AReinforcement learning is learning what to do—how to map situations to actions—so as to maximize a numerical reward signal.
- BReinforcement learning is learning what to do—how to map situations to actions—so as to minimixe a numerical reward signal.
- Cboth
2Supervised learning
- Alabeled samples provided by an external expert
- Bfinding an underlying structure in unlabeled data
- CLearn how to achieve a goal
3labeled samples provided by an external expert
- AUnsupervised learning
- BSupervised learning
- CReinforcement Learning
4 Supervised learning
- AClassification
Regression - BRegression
- Cboth
- AClassification
5Clustering
- ASupervised learning:
- BUnsupervised learning:
- CReinforcement Learning
6Learn how to achieve a goal
- AReinforcement Learning
- BSupervised learning:
- CUnsupervised learning
7the actor that performs actions on the environment, following a policy.
- AAgent
- BEnvironment
- CState
8Environment
- AThe context in which the agent perform actions
- Bthe actor that performs actions on the environment, following a policy.
- CA particular combination of state features and values
9A particular combination of state features and values
- AState
- BAction
- CPolicy
10Action
- AA particular operation that the agent carries out in the environment
- BA particular combination of state features and values
- Cthe actor that performs actions on the environment, following a policy.
11The value to be maximized by the agent. the primary basis for adapting the policy.
- AReward signa
- BValue function
- CPolicy
12Value function
- ASpecifies the value of a state: the reward an agent can expect to accumulate over the future, starting from that state
- BThe behavior of the agent. The mapping from perceived states of the environment to actions to be taken when in those states
- Cboth
13The behavior of the agent. The mapping from perceived states of the environment to actions to be taken when in those states
- APolicy
- BValue function
- CReward signal
14 the core of a reinforcement learning agent in the sense that it alone is sufficient to determine behavior
- APOLICY
- BVALUE FUNCTION
- CACTION
15for a particular state if we execute the same action, we transition to the same next state
- ADeterministic environment
- BStochastic environment
- Cmodel-based
16Stochastic environment
- Afor a particular state if we execute the same action, we transition to a next state based on a probabilities distribution
- Bfor a particular state if we execute the same action, we transition to the same next state
- Cfor a particular state if we execute the same action, we transition to a next state based with out a probabilities distribution
17Which of the following is an example of a deterministic environment?
- AChess
- BRoll the dice
- CRacing against other drivers
- DPlaying golf via trial and error
18In a stochastic environment, what happens when you execute the same action in the same state?
- AIt always leads to the same next state.
- BIt leads to a next state based on a probability distribution.
- CThe universe explodes!
- DA talking parrot appears and gives you advice.
19What is a model in the context of reinforcement learning?
- AA famous actor who gives advice on how to win games.
- BA representation of the environment.
- CA fancy car you drive in racing games.
- DA catchy jingle you use to remember game strategies.
20Model-based methods in reinforcement learning:
- ARequire no prior knowledge of the environment.
- BPredict the next state based on the current state and action.
- CAre only useful for playing golf.
- DInvolve playing chess without knowing the rules.
21Model-free methods in reinforcement learning:
- AStart with a blank sheet and collect experiences.
- BAlways require a detailed model of the environment.
- CAre only useful for rolling dice
- DInvolve making random guesses and hoping for the best.
22Which method can find "creative solutions" by not being constrained by a model?
- AModel-based
- BModel-free
- CModel airplane building
- DModel runway fashion show
23What's a possible advantage of using both model-based and model-free methods together?
- AIt makes learning more boring.
- BIt's a guaranteed way to win at chess
- CIt can lead to quicker solutions with less resource use.
- DIt confuses the traffic light at the intersection.
24What is the Markov property in a Markov Decision Process (MDP)?
- AThe future state depends only on past states.
- BThe future state depends on the current state and all past states.
- CThe future state depends only on the current state.
- DThe future state depends on the phase of the moon.
25In a Markov Decision Process (MDP), what does the 4-tuple (S, A, Pa, Ra) represent?
- AA recipe for cooking dinner
- BThe coordinates of a treasure map
- CThe components of a decision-making model
- DThe lyrics of a famous song
26What is the purpose of using the discount factor (ɣ) in MDPs?
- ATo make rewards larger
- BTo make future rewards weigh less in the decision-making process
- CTo increase the uncertainty in future rewards
- DTo buy discounted movie tickets
27What is the goal of an RL algorithm in a Markov Decision Process?
- ATo find the optimal recipe for cookies
b) To maximize the expected return by finding the optimal policy
c) To collect as many seashells as possible
d) To predict the weather for next week - BTo maximize the expected return by finding the optimal policy
- CTo collect as many seashells as possible
- DTo predict the weather for next weeK
- ATo find the optimal recipe for cookies
28What is the key difference between a deterministic policy and a stochastic policy?
- AStochastic policies only work on odd-numbered days.
- BStochastic policies are easier to predict.
- CDeterministic policies always lead to the same action in the same state.
- DDeterministic policies are always better than stochastic policies.
29What does the state value function (vπ(s)) in MDPs indicate?
- AHow good a policy is for taking actions in a state.
- BThe expected return starting in a state and following a policy.
- CThe color of a traffic light.
- DThe number of cookies in the jar.
30What does the state-action value function (qπ(s,a)) in MDPs represent
- AThe cost of taking an action in a state.
- BThe weight of a decision in a policy.
- CHow good it is to take a specific action in a specific state.
- DThe length of a fishing rod.
31¿Qué es una política óptima en un MDP?
- AUna política que siempre toma la misma acción en el mismo estado.
- BUna política que maximiza la recompensa final para todos los estados.
- CUna política que cambia constantemente para mantener al agente en suspenso.
- DUna política que solo funciona los miércoles.
32¿Cuál es el propósito de usar el factor de descuento (ɣ) en MDPs?
- APara aumentar el tamaño de las recompensas.
- BPara reducir el peso de las recompensas futuras en la toma de decisiones.
- CPara comprar boletos de película a mitad de precio.
- DPara hacer que las recompensas futuras sean más inciertas.
33¿Qué representan las ecuaciones de optimización de Bellman en la resolución de MDPs?
- ARecetas de cocina para platos exquisitos.
- BUna ruta hacia la iluminación espiritual.
- CUna forma de encontrar la política óptima y resolver problemas de RL.
- DUn acertijo matemático sin solución.
34¿Qué tipo de políticas promueven la exploración y la explotación en RL?
- APolíticas determinísticas.
- BPolíticas ε-greedy.
- CPolíticas aleatorias.
- DPolíticas que solo funcionan en días impares.
35¿Cuál es la diferencia entre recompensas "sparse" y recompensas "dense" en RL?
- ALas recompensas sparse son más densas en información.
- BLas recompensas dense se otorgan más puntualmente, mientras que las sparse son más frecuentes.
- CLas recompensas sparse son más abundantes en el universo.
- DLas recompensas dense son más ricas en sabor.
36¿Qué significa SARSA en el contexto de RL?
- ASaber, Aprender, Resolver, Salvar, Avanzar
- BState, Action, Reward, State', Action'
- CSimular, Analizar, Resolver, Saltar, Avanzar
- DSúper Agent, Robot, Salto, Aventura
37¿Por qué es importante diseñar las recompensas cuidadosamente en RL?
- APara que el agente siempre elija la acción más aburrida.
- BPara evitar que el agente desarrolle estrategias perversas que maximicen la recompensa final sin considerar otros aspectos.
- CPorque las recompensas son irrelevantes en RL.
- DPara confundir al agente y hacer que haga cosas extrañas.
38¿Por qué se llama a TD-Learning "semigradiente" en el contexto del aprendizaje por refuerzo?
- APorque utiliza solo la mitad de los datos de entrenamiento.
- BPorque se enfoca en el cálculo de la derivada de la función de valor.
- CPorque involucra el uso de "bootstrapping" para estimar V(St) con una estimación previa.
- DPorque solo se aplica en problemas matemáticos a medias.
39 Which of the following options is a technique used as an approximation function in machine learning?
- ADecision trees
- BVacation planning
- CCooking recipes
- DLullabies
40What is Gradient Descent?
- AA popular cooking recipe.
- BAn optimization algorithm used to minimize a target/objective function.
- CA board game.
- DA machine learning technique for creating neural networks.
41Which of the following statements is true about Gradient Descent?
- AIt is an algorithm that always finds the optimal solution.
- BIt has very low computational cost.
- CIt starts from a random point on a function and moves down its slope.
- DIt is only used in foreign language learning.
42Why is Stochastic Gradient Descent (SGD) preferred over Gradient Descent in some cases?
- ABecause SGD is slower.
- BBecause SGD uses the entire dataset in each iteration.
- CBecause SGD greatly reduces computations by choosing a random data point in each iteration.
- DBecause Gradient Descent is not suitable for machine learning.
43 What is an important feature of Stochastic Gradient Descent (SGD)?
- AIt is computationally expensive.
- BIt only works with neural networks.
- CIt chooses a random data point in each iteration.
- DIt always selects all data in each iteration.
44What does Gradient Descent do in the optimization process?
- AIncreases the objective function.
- BMoves upward on the function's slope.
- CMinimizes an objective function by moving downward on its slope in successive iterations.
- DStays in the same place without moving.
Informationen
- Beschreibung
- Quiz am Ohne Titel, erstellt von Naia Landaluce Marquínez am 03/01/2024.
- Sprache
- Español
- Aufrufe
- 0
- Bewertung
- Noch keine Bewertungen Bewerte dieses Quiz