18
votes
Accepted
Water Jug Problem in AI
Recoverable, as we can pour all water back to the 12L jug to restore the original state, hence any state derived thereof (by following the same steps from the start). The problem is, also, solvable:
<...
10
votes
In what ways can we distinguish between a human and bot behavior?
The most common/obvious way is a challenge-response test that is easy for humans but hard for computers (of course, but not only, CAPTCHA).
This kind of test is very effective{1} but falls under the ...
- 1,997
8
votes
Accepted
In principle, what is the relation between Artifical Intelligence and Turing machine?
Turing machines are a model of computation, one way of formally defining the concept of an algorithm. While Turing machines are usually defined using barebones input/output capabilities, it is not ...
- 274k
8
votes
Accepted
How to determine the time and memory complexity for solving a sliding-tile puzzle?
The complexity of the BFS and DFS algorithms depend heavily on the graph being analyzed, and the search strategy being used. If we have a method to consistently get "closer" to a solution, ...
- 290
7
votes
Google DeepDream Elaborated
The idea of DeepDream is this: pick some layer from the network (usually a convolutional layer), pass the starting image through the network to extract features at the chosen layer, set the gradient ...
- 640
7
votes
Accepted
Bidirectional Dijkstra vs Dijkstra
Absolutely yes! your arguments are correct. And, as matter of fact, it is very easy to come up with a graph where Bidirectional Dijkstra would expand more nodes than Unidirectional Dijkstra, following ...
- 3,433
7
votes
Does the the undecidability of the Halting Problem eliminate the possibility of 'Hard AI'?
You're understanding of "solve all instances of the Halting Problem" is flawed. All instances means ALL instances. Every single one, ever.
For example, no human knows if a Turing Machine verifying ...
- 29.6k
7
votes
What's the input to the decoder in a sequence to sequence autoencoder?
I was wondering the same and just stumbled across a nice tutorial by Quoc V. Le. The following explanation deals with the conditional case since this seems to be the common case. My explanation is ...
- 171
7
votes
Accepted
Why do we use the log in gradient-based reinforcement algorithms?
We often take the logarithm because:
Maximizing $\log \Phi(x)$ is equivalent to maximizing $\Phi(x)$, so in maximum-likelihood problems, we can maximize the log of the likelihood instead of ...
D.W.♦
- 154k
7
votes
Water Jug Problem in AI
I couldn't in a quick Google through the textbooks tell whether recoverability requires immediate, single-move recoverability (ctrl-Z-style) with each undo step being O(1) complexity, and which can be ...
- 200
6
votes
Comparison between IDA* and Recursive best first search
Let me please start by succintly summarizing the behaviour of RBFS. For a thorough explanation of the algorithm refer to the original journal paper: Richard Korf. Linear-space best-first search. ...
- 3,433
6
votes
The meaning of discount factor on reinforcement learning
The discount factor does not represent the likelihood of reaching the state $s′ $from the state $s$. That would be $p(s'|s,a)$, which is not used in Q-Learning, since it is model-free (only model-...
- 471
6
votes
Accepted
Why does an admissible heuristic mean A* is optimal?
With an admissible heuristic
The heuristic defines which nodes will be explored first, but does not change the final path found.
In your example, the heuristic will cause the path to Z to be ...
- 306
6
votes
Accepted
How do neural networks create results like its inputs?
These are known as Autoencoders. As you said, these neural networks are trained to produce output that is similar to the input, rather than output a classification of some kind.
Internally, they do ...
- 368
6
votes
Is the halting problem claim true with the advent of AI
Even if we can design better AI solving problems we did not think possible even 10 years ago, they will always be computer programs. And as you said it yourself, "Alan Turing states that, there can't ...
- 881
6
votes
How does machine learning relate to artificial intelligence?
Machine Learning is a subset of (the scientific field of) Artificial Intelligence.
What is ML?
Machine Learning is defined by Tom Mitchel:
A computer program is said to learn from experience E ...
- 2,350
6
votes
Artificial intelligence and undecidibility
No. The Post correspondence problem is undecidable. That means that no computer program can solve it (in all cases). "Artificial intelligence" is just a computer program. Actually, "artificial ...
D.W.♦
- 154k
6
votes
Accepted
Is Artificial General Intelligence possible with our current machine learning models?
The short answer is, we don't know! This is an open question in AI research.
We know how neurons transmit signals, and can simulate that in a straightforward way: that's how layered perceptron models ...
- 7,038
6
votes
Water Jug Problem in AI
Suppose the leftmost jug had 3 liters, the center jug had 2 liters, the right jug is empty. Suppose you poured the leftmost jug to the center jug. Now the leftmost jug is empty, and the center jug ...
- 274k
5
votes
Accepted
What is the difference between Cased-based Reasoning and Rule-based reasoning?
As said in the comments, there may be a part of subjectivity in the answers to this question.
Yet, I think it is fair to say that case-based reasoning mostly belongs to what is often called ...
- 348
5
votes
Accepted
Turing tests and humans
The basic objective of a Turing test is to come up with such an intelligent machine, which actually mimics the human way of communicating.
For that purpose, we'll carefully need to think about what ...
- 207
5
votes
How are artificial intelligence and Natural intelligence compared?
You are probably referring to ConceptNet (created by the Massachusetts Institute of Technology under its Artificial Intelligence program).
It's a software system / semantic network containing lots of ...
- 1,997
5
votes
Prove consistency of maximum of two consistent heuristic functions?
Proof (Show consistency property of $h_3$):
$$
h_3(n) = \max(h_1(n), h_2(n)) \\
\leq max(h_1(n')+c(n,a,n'), \ h_2(n')+c(n,a,n')) \\
\leq \max(h_1(n'), \ h_2(n')) + c(n,a,n') = h_3(n') + c(n,a,n')
$$...
- 558
5
votes
Accepted
Explanation of proof of why connectedness is not conjunctively local of any order $k$
The predicate $\varphi_0$ depends on at most $k$ points. There are $k+1$ middle squares. So $\varphi_0$ cannot depend on all of them. That is, there is a middle square that $\varphi_0$ does not depend ...
- 274k
5
votes
Accepted
How did the Logic Theorist prove the Pons Asinorum?
If you google "logic theorist source code" you find this which is clearly not the original source code, but presumably is a modernization of the ideas in the code. You can also find this 1963 RAND ...
5
votes
Being stuck and frustrated with my masters project
I have been implementing a branch and bound solver with heuristics for an NP-hard problem. It got complicated at some points and had to reimplement parts a couple of times. The problem was (I think), ...
- 4,458
5
votes
Accepted
What is the purpose of learning propositional logic
Briefly, one can't learn AI without it.
Some of the main content areas of AI are based on logic: knowledge representation, planning, natural language semantics and reasoning. Have a look at the ...
- 921
4
votes
Accepted
Exploring and interpolating a function using machine-learning?
I would look into the field of "optimal experimental design" in bayesian inverse problems, particularly the recent work of Alen Alexandrian.
http://arxiv.org/abs/1410.5899
http://www4.ncsu.edu/~...
- 1,049
4
votes
that there would be no perfect strategy for poker
Nash proved that every zero-sum game has a perfect strategy called a Nash equilibrium. This is a possibly randomized strategy for both players such that given that one player follows the strategy, it ...
- 274k
Only top scored, non community-wiki answers of a minimum length are eligible