28
Here are three survey papers that examine the use of machine learning in time series forecasting:
"An Empirical Comparison of Machine Learning Models for Time Series Forecasting" by Ahmed, Atiya, El Gayar, and El-shishiny provides an empirical comparison of several machine learning algorithms, including:
"...multilayer perceptron, Bayesian neural ...
21
You will need some discrete mathematics. Graphs, trees, and so forth. These are the structures underlying AI.
You will need some programming skills, especially in languages such as Prolog and LISP.
A lot of AI systems are programmed in these languages.
You will need some logic. Propositional and predicate calculus. Their syntax and semantics. Perhaps some ...
18
There is no "official Turing test" so there's no concept of "officially pass[ing] the test". Turing described a methodology that one might use to evaluate artificial intelligences. The organizers of the event that Eugene Goostman won implemented that methodology in a particular way and the program satisfied the criteria the organizers had chosen. In that ...
18
The Church-Turing thesis says that the informal notion of an algorithm as a sequence of instructions coincides with Turing machines. Equivalently, it says that any reasonable model of computation has the same power as Turing machines.
An artificial intelligence is a computer program, i.e., an algorithm. If the Church-Turing thesis holds, then you could ...
17
Neural Networks are not magic. If you treat them like they are and just throw data at them without thinking you're going to have a very bad time.
You need to stop and ask youself "Is milliseconds since 1970 actually going to be predictive of the event I'm interested in?" The answer you should arrive at immediately is no. Why? For every instance you ...
16
The whole state space for chess is enormous - it can be roughly estimated as 1043 (Shannon number (Shannon, 1950), (Wikipedia)).
The idea you present - Reinforcement Learning agents playing with each other to learn the game - was successfully applied to Backgammon - TD-Gammon (Tesauro, 1995), (Chapter in Reinforcement Learning by Sutton&Barto). It also ...
16
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:
[12, 0, 0]
[4, 8, 0]
[4, 3, 5]
[9, 3, 0]
[9, 0, 3]
[1, 8, 3]
[1, 6, 5]
[6, 6, 0] # <-- SOLVED; 7 steps
DETAILS: The problem stated as-is is ...
15
The halting problem is not a statement about intelligence (human or artificial) it is a statement about the limits of mathematics. It is an historically important example of an undecidable problem.
An artificial (or human) intelligence can certainly look for and find many sorts of different infinite loops in real programs. And the halting problem doesn't ...
12
Yes it can, and has been. In the paper Map-Reduce for Machine Learning on Multicore they discuss using the Map-Reduce paradigm for several common ML algorithms including ANNs.
12
If the heuristic function is not admissible, than we can have an estimation that is bigger than the actual path cost from some node to a goal node. If this higher path cost estimation is on the least cost path (that we are searching for), the algorithm will not explore it and it may find another (not least cost) path to the goal.
Look at this simple example....
12
This is an expansion of this presentation.
Because the state graph consists of two disconnected components of equal size.
Without loss of generality we can assume that the target state is $1\;2\;3\;...\;15\;\Box$.
1 2 3 4
5 6 7 8
9 10 11 12
13 14 15 *
Given a state $S$ a permutation inversion is a tile $T_i$ that is placed after $T_j$ but $i &...
11
So it seems you are intrigued about the relationship between the informedness of a heuristic function and its pruning power.
This is a well-known relationship established in the literature from the 80s (see for example Pearl, Judea. Heuristics, Addison-Wesley, 1984 who, by the way, has been awarded this year with the Alan Turing award).
As you already ...
10
Although Alan Turing is of course a very important computer scientist, the Turing test is only superficially related to computer science. It is more related to philosophy.
As long as machines exist people have wondered whether it is possible to construct a machine that can think. But to answer this question we must first answer the question what it means -- ...
9
The thing is that with simultaneous moves, the optimal strategy is harder to guess, because you need to compute something that is not always obviously winning.
Have a look at Nash equilibrium and the prisoner's dilemma if you don't know them yet. This is the kind of reasoning that you will need each time you are considering two simultaneous moves, instead ...
9
As noted by Thomas Klimpel in the comments, a certain acceptance probability is often used, which is equal to say $0.8$. The following is a simple iterative method to find a suitable initial temperature, proposed by Ben-Ameur in 2004 [1]. In the following, $t$ is a strictly positive transition, $\max_t$ and $\min_t$ are the states after and before the ...
9
For the field of A.I. and machine learning, I would recommend you to explore and learn more about these topics:
Statistics
Probability
Stochastic processes
Bayesian Data Analysis
Convex Optimization
Graph Theory
With your math background, you could easily pick any good machine learning book and learn the required math that you don't have as you go. Kevin ...
9
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 HIP (Human Interactive Proofs) area: it's not transparent.
Typical, "simple" approaches to distinguish human website traffic from Bot are:
time it takes to ...
9
I think the prizes you're referring to are the Loebner Prize. According to the Wikipedia page (see prior link), the winner for 2014 is 'Rose' by Bruce Wilcox. That program did not win one of the one-time-only prizes, but did get $4,000 in prize money. 'Eugene Goostman' competed in 2005 and 2008, finishing second both times.
The competition 'Eugene ...
8
Perhaps the simplest way to understand the proof is by the idea of a conserved quantity : find some quantity that can be derived from a configuration and show that every move preserves that quantity. A static version of the idea is found in the following old puzzle:
Remove the northeast and southwest corner squares from a standard 8x8 checkerboard. Can ...
8
With neural networks, you always need to randomly initialize your weights to break symmetry.
If you don't use a non-linear activation function in the hidden units, then you might as well have stayed with a single layer. Your network is now just a composition of two linear functions, which is of course just another linear function.
That learning rate seems ...
8
Roughly speaking, over-fitting typically occurs when the ratio $\frac{\text{complexity of the model}}{\text{training set size}} $ is too high. Think of over-fitting as a situation where your model learn the training data by heart instead of learning the big pictures which prevent it from being able to generalized to the test data: this happens when the model ...
8
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 difficult to extend them so that they can apply to the kind of problems that AI is after.
The relation between AI and Turing machines is thus:
Artificial ...
7
One very good resource is the Neural Network FAQ. The question depends a lot on your problem. If the problem is linear in nature, there is no reason to have any hidden layers. If the problem is non-linear, often a single hidden layer with around 10 hidden neurons will do the trick.
There is a very similar question (with a very similar answer) at ...
7
That list would be endless ... I will just try to provide a number of representative examples according to different criteria:
Best-first search (BFS): they are complete, i.e., they are guaranteed to find a solution provided that one exists and they are admissible, i.e., they are guaranteed to find the optimal solution provided, again, that one exists. ...
7
Back to the definitions:
$$H(L\mid A) = \sum_a p(A=a) H(L \mid A=a).$$
As you compute, $P(A=true)=6/8$ and $P(A=false)=2/8$.
However, you don't compute $H(L\mid A=true)$ but instead compute $P(L=positive\mid A=true)$. [and the same for $A=false$.].
With standard definition of $H()$ we get,
$$H(L\mid A=true) = - 4/6\log_2(4/6) - 2/6\log_2(2/6) = 0....
7
While Anton's answer is absolutely perfect let me try to provide an alternative answer: being admissible means that the heuristic does not overestimate the effort to reach the goal, i.e., $h(n) \leq h^*(n)$ for all $n$ in the state space (in the 8-puzzle, this means just for any permutation of the tiles and the goal you are currently considering) where $h^*(...
7
At a stretch it is an expert system (such as fuzzy logic). As you are not running an algorithm to perform feedback onto the decision parameters based on the output, it's not really learning. However, performing feedback is not the only indicator whether an alogirthm is AI. One could argue that if it acts in a way that appears intelligent, that's all that ...
7
The deterministic requirement isn't all that constraining. That just implies your vehicle is certain of the state it's in. That being said, you'll probably want to plan paths in a way that allows you to avoid obstacles. The best way I've seen this done is with sampling-based planners. Steven LaValle wrote the central academic resource on this topic: Planning ...
7
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 at that layer equal to the activations themselves, and then backpropagate to the image.
Why does it make sense? Intuitively, it amplifies the features that are ...
7
The state described (3, 2, 0) in the currently accepted answer is not achievable from the initial state of the given problem, unless "pour water on the ground" is a valid move, in which case I agree it's non-recoverable without a very good shop-vac.
I couldn't in a quick Google through the textbooks tell whether recoverability requires immediate, single-...
Only top voted, non community-wiki answers of a minimum length are eligible
