nbro
Moderator
  • Member for 5 years, 4 months
  • Last seen this week
What is self-supervised learning in machine learning?
Accepted answer
81 votes

Introduction The term self-supervised learning (SSL) has been used (sometimes differently) in different contexts and fields, such as representation learning [1], neural networks, robotics [2], natural ...

View answer
Are neural networks prone to catastrophic forgetting?
Accepted answer
65 votes

Yes, indeed, neural networks are very prone to catastrophic forgetting (or interference). Currently, this problem is often ignored because neural networks are mainly trained offline (sometimes called ...

View answer
Is a switch from R to Python worth it?
32 votes

Of course, this type of questions will also lead to primarily opinion-based answers. Nonetheless, it is possible to enumerate the strengths and weakness of each language, with respect to machine ...

View answer
How can an AI train itself if no one is telling it if its answer is correct or wrong?
31 votes

By "company A has a large human face database so that it can train its facial recognition program more efficiently" the article probably means that there is a training dataset $S$ of the form $$ S = ...

View answer
Is there any research on the development of attacks against artificial intelligence systems?
Accepted answer
28 votes

Yes, there is some research on this topic, which can be called adversarial machine learning, which is more an experimental field. An adversarial example is an input similar to the ones used to train ...

View answer
What is the credit assignment problem?
Accepted answer
26 votes

In reinforcement learning (RL), an agent interacts with an environment in time steps. On each time step, the agent takes an action in a certain state and the environment emits a percept or perception, ...

View answer
Where can I find the proof of the universal approximation theorem?
Accepted answer
20 votes

There are multiple papers on the topic because there have been multiple attempts to prove that neural networks are universal (i.e. they can approximate any continuous function) from slightly different ...

View answer
What is the difference between First-Visit Monte-Carlo and Every-Visit Monte-Carlo Policy Evaluation?
Accepted answer
20 votes

The first-visit and the every-visit Monte-Carlo (MC) algorithms are both used to solve the prediction problem (or, also called, "evaluation problem"), that is, the problem of estimating the value ...

View answer
What are some well-known problems where neural networks don't do very well?
19 votes

In theory, most neural networks can approximate any continuous function on compact subsets of $\mathbb{R}^n$, provided that the activation functions satisfy certain mild conditions. This is known as ...

View answer
What's the difference between model-free and model-based reinforcement learning?
19 votes

In reinforcement learning (RL), there is an agent which interacts with an environment (in time steps). At each time step, the agent decides and executes an action, $a$, on an environment, and the ...

View answer
Why do we need common sense in AI?
Accepted answer
18 votes

Commonsense knowledge is the collection of premises that everyone, in a certain context (hence common sense knowledge might be a function of the context), takes for granted. There would exist a lot of ...

View answer
What are the differences between Q-Learning and A*?
Accepted answer
16 votes

Q-learning and A* can both be viewed as search algorithms, but, apart from that, they are not very similar. Q-learning is a reinforcement learning algorithm, i.e. an algorithm that attempts to find a ...

View answer
What are examples of promising AI/ML techniques that are computationally intractable?
16 votes

AIXI is a Bayesian, non-Markov, reinforcement learning and artificial general intelligence agent that is incomputable, given the involved incomputable Kolmogorov complexity. However, there are ...

View answer
Why do most deep learning papers not include an implementation?
15 votes

The paper's authors needed to implement their models anyway in order to conduct their experimentations, so why not publish the implementation? Some papers and authors actually provide a link to their ...

View answer
How does one prove comprehension in machines?
14 votes

This is one of the most important issues in the philosophy of artificial intelligence. The most famous philosophical argument that attempts to address this issue is the Chinese Room argument ...

View answer
Problems that only humans will ever be able to solve
Accepted answer
14 votes

Informally, AI-complete problems are the most difficult problems for an AI. The concept is not mathematically defined yet, as e.g. NP-complete problems. However, intuitively, these are the problems ...

View answer
What sort of mathematical problems are there in AI that people are working on?
Accepted answer
13 votes

In artificial intelligence (sometimes called machine intelligence or computational intelligence), there are several problems that are based on mathematical topics, especially optimization, statistics, ...

View answer
What are examples of promising AI/ML techniques that are computationally intractable?
13 votes

Exact Bayesian inference is (often) intractable (i.e. there is no closed-form solution, or numerical approximations are also computationally expensive) because it involves the computation of an ...

View answer
Should neural nets be deeper the more complex the learning problem is?
13 votes

Deeper models can have advantages (in certain cases) Most people will answer "yes" to your question, see e.g. Why are neural networks becoming deeper, but not wider? and Why do deep neural networks ...

View answer
What algorithms are considered reinforcement learning algorithms?
Accepted answer
12 votes

The dynamic programming algorithms (like policy iteration and value iteration) are often presented in the context of reinforcement learning (in particular, in the book Reinforcement Learning: An ...

View answer
What is the difference between active learning and online learning?
Accepted answer
12 votes

Active learning (AL) is a weakly supervised learning (WSL) technique where you can have both labelled and unlabelled data [1]. The main idea behind AL is that the learner (or learning algorithm) can ...

View answer
Is there a fundamental difference between an environment being stochastic and being partially observable?
Accepted answer
11 votes

I think the distinction is made more for conceptual reasons, which has practical implications, so let me review the usual definitions of a stochastic and partially observable environment. A stochastic ...

View answer
How should the neural network deal with unexpected inputs?
Accepted answer
11 votes

This is a very important problem that is usually overlooked. In fact, when training a neural network, there's often the implicit assumption that the data is independent and identically distributed, i....

View answer
How is BERT different from the original transformer architecture?
Accepted answer
11 votes

What is a transformer? The original transformer, proposed in the paper Attention is all you need (2017), is an encoder-decoder-based neural network that is mainly characterized by the use of the so-...

View answer
How is iterative deepening A* better than A*?
Accepted answer
11 votes

A* is a best-first search algorithm, which means that it is an algorithm that uses both "past knowledge", gathered while exploring the search space, denoted by $g(n)$, and an admissible heuristic ...

View answer
Can reinforcement learning be used for tasks where only one final reward is received?
Accepted answer
10 votes

RL can be used for cases where you have sparse rewards (i.e. at almost every step all rewards are zero), but, in such a setting, the experience the agent receives during the trajectory does not ...

View answer
What is the "Hello World" problem of Reinforcement Learning?
10 votes

MNIST (along with CIFAR) may be the "Hello World" of supervised learning for image classification, but it is definitely not the "Hello World" of all machine learning techniques, ...

View answer
What is a recurrent neural network?
10 votes

A recurrent neural network (RNN) is an artificial neural network that contains backward or self-connections, as opposed to just having forward connections, like in a feed-forward neural network (FFNN)....

View answer
What are some online courses on artificial general intelligence?
Accepted answer
9 votes

As far as I know, no AGI system has yet been created, so that's why there aren't yet many courses on AGI. However, there are a few courses that attempt to address AGI as the main topic but from ...

View answer
What is the difference between a stochastic and a deterministic policy?
Accepted answer
9 votes

A deterministic policy is a function of the form $\pi_{\mathbb{d}}: S \rightarrow A$, that is, a function from the set of states of the environment, $S$, to the set of actions, $A$. The subscript $_{\...

View answer
1
2 3 4 5
25