Questions tagged [machine-learning]

Theoretical questions about Machine learning, especially Computational Learning Theory, including Algorithmic Learning Theory, PAC learning, and Bayesian Inference

Filter by
Sorted by
Tagged with
4
votes
2answers
378 views

Can you make a different learner?

My questions are: [Solved by Dave] Given a learner N, can you design a learner M that behaves differently from N? No. and [Solved by Dave] Given a learner N, can you design a learner M that is more ...
4
votes
1answer
236 views

Adversarial Machine Learning, Learning with (Malicious) noise

I am reading some old papers regarding Learning With Malicious Noise. In one of them, Learning in the presence of Malicious Errors, by Kearns and Li $[1]$ (https://www.cis.upenn.edu/~mkearns/papers/...
4
votes
1answer
610 views

Sample complexity of distinguishing two Gaussian distributions?

Below is a description of the problem: Suppose I have two $p$-dimensional Gaussian distributions with the same covariance matrix $\Sigma$ and means $\mu_1$, $\mu_0$. And I can get $n$ samples $X_1^{(...
4
votes
1answer
951 views

Are (empirical) Rademacher complexity always positive?

Rademacher complexity and empirical Rademacher complexity are used to provide upper bound on the loss of solving an learning problem. That seems to imply that Rademacher complexity and empirical ...
4
votes
2answers
156 views

Are there hypothesis classes that are hard to learn but easy to test?

Let $H$ be a binary hypothesis class, it is easy to see that if $H$ is (efficiently) properly PAC learnable then it is also (efficiently) testable (here we use the standard notion of within or $\...
4
votes
1answer
1k views

Neural Networks to classify accelerometer double taps

I'm building an application for Android devices that requires it to recognize, by accelerometer data, the difference between walking noise and double tapping it. I'm trying to solve this problem using ...
4
votes
1answer
113 views

“Learning” when test and train distributions don't match

We know that the theory of PAC-learning is distribution-free, i.e. assuming that the test and train distributions are the same, we have guarantees on learning the hypothesis. Question: what if the ...
4
votes
2answers
878 views

Are single hidden-layered neural networks at least as good as multi hidden-layered neural networks?

If I have a multi hidden-layered neural network that is getting a better approximation for a function than a single one, does that mean that there is something "fishy" about my multi layered one ...
4
votes
1answer
1k views

Stochastic Gradient Descent with integer arithmetics

Most implementations of stochastic gradient descent (SGD) rely on floating points. Is there implementations using infinite or finite precision integer arithmetics ?
4
votes
1answer
79 views

Terminology and references for a learning model

Let's say we're doing regression over $[0,1]^d$ -- either in the PAC sense with bounded-range agnostic noise or in the more classical-statistics sense with additive Gaussian noise. Suppose further ...
4
votes
1answer
648 views

Autoencoders and information compression

Disclaimer: I know very (very) little about deep nets, besides what an introductory course on machine learning would teach on neural networks, and skimming some paper abstracts and introductions. If ...
4
votes
1answer
256 views

Are there any learning algorithms with any provable guarantees for manifold learning or manifold regularization?

First of all, I want to make clear that my question is about algorithms. I'd like to know if there are any algorithms with provable guarantees in the context of manifold learning (or manifold ...
4
votes
1answer
708 views

Using Indicator Functions as Transfer Functions for Neural Networks

Does there exist any theory (other than Cybenko's proof of the Universal Approximation Theorem with sigmoids) advocating the use of indicator functions as transfer functions for machine learning with ...
4
votes
1answer
1k views

VC dimension of intersection of half-spaces

Define $$l_i(x) := \text{sgn} \left( w_i^\top x - b_i \right)$$ for $i=1,...,n$, where $x \in \mathbb{R}^d$. Then define the classifier $$ g(x) := \max \{ l_1(x),..., l_n(x) \}$$ which represents ...
4
votes
1answer
196 views

Active learning for inferring a convex optimization formulation

I was wondering if anybody knows of any relevant references on the general topic of active learning for gradually inferring/updating a convex opt. formulation. As a specific example, I am thinking of ...
4
votes
1answer
255 views

Higher-order and black-box clustering

As far as I understand a large number of clustering problems can be formulated as: $\underset{\textbf{P}}{ \text{argmin}} \; \sum_{i,j} f \left(x_i, x_j\right)$ where $\textbf{P}$ is a partitioning ...
4
votes
1answer
128 views

Are there perceptrons that maximize the average margin, rather than the minimum?

A perceptron is a linear classifier. The standard method for training a perceptron involves maximizing the minimum margin. That is, we are trying to find: $x^* = \text{argmax}_{x\in \text{unit ...
4
votes
0answers
71 views

Status of the Junta Problem (soft question)

Does the learning theory community in general believe that juntas can be learned in polynomial time? The naive algorithm works in quasi-polynomial time. MOS's paper shows how to solve the junta ...
4
votes
2answers
540 views

Universal Approximation - Are ReLUs discriminatory?

In Cybenko's elegant proof of the Universal Approximation Theorem (UAT) he proves that single hidden layer neural networks (with linear output layer) are universal approximators whenever their ...
4
votes
0answers
109 views

Learning hidden variable distribution

Consider a set of $k$ continuous variables. Each variable $x_k$ is associated with a hidden distribution from which its value is sampled independently of other variables. I am given a set of ...
4
votes
0answers
1k views

Universal Approximation Theorem for non-sigmoidal activation functions

The most cited Universal Approximation Theories for multi-layer feedforward neural networks by Cybenko (1989) and Hornik (1991) assume the activation functions of the network to be sigmoidal. However, ...
4
votes
1answer
181 views

Kleinberg-consistency of spectral clustering

Spectral clustering refers to a family of graph-based algorithms, which usually rely on a similarity function rather than a metric, though a metric $\rho(x,y)$ can always be converted to a similarity ...
4
votes
0answers
98 views

Adversarial distributions for PAC lower bounds

The various PAC lower bounds (realizable, agnostic, bounded noise) construct distributions supported on $d$ points, where $d$ is the VC-dimension of the hypothesis class in question. Does anyone ...
4
votes
0answers
532 views

Practical Implications of Kolmogorov's Result on the Universal Approximation Theorem with Neural Networks

After having read matus's beautiful answer in this thread explaining (among other things) Kolmogorov's result regarding the Universal Approximation Theorem with Neural Networks, I wonder: if just $\...
3
votes
1answer
295 views

Does Approx Carathéodory's theorem implies dimensionality reduction

Carathéodory's theorem says that if a point $x$ of $R^d$ lies in the convex hull of a point set $P$, then there is a subset $P′ \subseteq P$ consisting of $d + 1$ or fewer points such that $x$ can be ...
3
votes
1answer
188 views

Is this variant of PAC learning known?

Here is a problem I've never seen, in a model similar to the PAC model. It asks a similar question to PAC learning, but wishes to optimize, rather than learn. I wonder if this problem is known, has ...
3
votes
1answer
2k views

Computational complexity of classifying with an already-trained SVM

If I have a support vector machine which has already been trained, what is the computational complexity of classifying a new example using that machine? I care about both time and space complexity. ...
3
votes
1answer
455 views

Online to batch sample complexity

It is well known that a mistake bound can be converted to a PAC bound. I know how to prove a sample complexity of $$ O( (1/\epsilon)[M + \log(M/\delta)] ), $$ where $M$ is an upper bound on the number ...
3
votes
2answers
238 views

Discerning the best model for a problem

This is a vague question. I will do my best, I think it has definite answers. I am hoping for answers of the form "Read book x, learn this specific topic, read this paper/s". What is bothering me is ...
3
votes
2answers
434 views

PAC-learning bound with epsilon-cover of hypothesis class

In this video at 43:00, a version of the PAC bound for generalization error $\epsilon$, which I hadn't seen before, is quoted: $$\epsilon^2 < \frac{\log{|H_\epsilon|} + \log{1/\delta}}{2m}$$ ...
3
votes
1answer
766 views

Objective function for stochastic optimization

Stochastic Optimization problems in general deals with random variables in the 'loss function'. Incase of a Deterministic optimization problem with basic objective $\parallel Ax-b \parallel_2^2$, we ...
3
votes
1answer
114 views

Is there an algorithm that's “like” cross-validation for approximation algorithms of NP-hard problems?

I normally do machine learning work, and when I'm evaluating an algorithm on a data set, I always use cross-validation to determine how effective the algorithm is. Is there a similar method for ...
3
votes
2answers
205 views

How can AIC converge in the limit when even 2 parameter models can have infinite VC dimension?

AIC-based model-selection converges to zero error in the limit, and also has finite-sample convergence that is rate-optimal with respect to worst case minimax error [1]. (Note that AIC refers to ...
3
votes
1answer
275 views

Rademacher complexity beyond the agnostic setting

The way I know of to bound generalization error by Rademacher complexity is Theorem 2.4 in this lecture notes, http://ttic.uchicago.edu/~tewari/lectures/lecture9.pdf. Here the quantity on the LHS that ...
3
votes
1answer
221 views

Agnostic Learning of K-Juntas over “small” distribution

I have two questions related to agnostic learning, one specific and one more general, specifically when the distribution relative to which the learner must operate is given explicitly as part of the ...
3
votes
1answer
78 views

Agnostic query learning for DFAs

Angluin's membership+equivalence query algorithm allows to efficiently and exactly learn a target $n$-state DFA. But what if the target DFA is huge, or the target concept is not even a regular ...
3
votes
1answer
144 views

Is there a gap between weak learning and PAC-learning?

For concreteness lets use the definitions of PAC and weak-learning as in the notes of Avrim Blum (http://www.cs.cmu.edu/~avrim/ML12/lect0208.txt) and also his notes on SQ-Learning (http://www.cs.cmu....
3
votes
1answer
153 views

Tighter Probability Bounds

Let $\mathcal{F}$ be a class of binary functions on a probability space $\Omega$. For $f \in \mathcal{F}$, let $P(f) =\mathbb{E}(f(Z))$ and $P_n(f) = \frac{1}{n} \sum_{i=1}^n f(Z_i)$ where $Z_i$'s are ...
3
votes
1answer
734 views

Examples of Fat-Shattering Dimension

What are some good examples for analysis of a class's Fat-Shattering dimension? By (Alon et al) I know that the Fat-Shattering Dimension characterizes the learnability of real-valued function classes ...
3
votes
2answers
407 views

Representation of procedural knowledge

I know that knowledge about relationships between things can be represented using ontologies and stored in some sort of file or database system. Can a network of procedural knowledge also be created ...
3
votes
1answer
235 views

How to deal with concept classes with exponential value of VC dimension

Let $C$ be a concept class with VC dimension $d$ exponential to the input size (i.e number of variables represented in each concept $c\in C$). I am looking for papers/resources/suggestions of how ...
3
votes
1answer
604 views

Applying Expectation Propagation to Factor Graph

Expectation Propagation(EP) is now quite a standard technique to approximate marginal in graphical model. Moreover, EP can replace sum-product algorithm in factor graph. For this reason, I try to ...
3
votes
1answer
137 views

Examples of nontrivial non-discriminatory functions

I am reading Cybenko's "Approximation by Superpositions of a Sigmoidal Function". The paper defines a discriminatory function as: $\sigma$ is discriminatory if for a measure $\mu$, \begin{align} \int ...
3
votes
0answers
105 views

Understanding Dudley Chaining Argument for Rademacher Bound

I follow the proof of the Dudley chaining/metric entropy bound of the (empirical) Rademacher complexity, but I don't have any intuition for why this bound should be true. In particular, I don't know ...
3
votes
0answers
177 views

Is this a known learning problem?

Let $(\mathcal{X},\rho)$ be a metric space (say, $\mathcal{X}=[0,1]$ with the Euclidean metric). Let $\alpha:\mathcal{X}\to[0,1]$ be unknown. Suppose that $\mathcal{X}$ is endowed with a distribution $...
3
votes
0answers
106 views

Do features always induce a metric?

It is well-known in functional analysis that an inner product always induces a norm and a norm always induces a metric, and the reverse directions do not hold in general. I am wondering if a similar ...
3
votes
0answers
65 views

Is there some research about infinitely many-armed bandit with non-stationary assumption?

Is there some research about infinitely many-armed bandit with non-stationary assumption? I have found the paper about infinitely many-armed bandit under stationary (or stochastic) assumption. And I ...
3
votes
1answer
134 views

Minimax agnostic risk for Lipschitz functions

For $L>0$, let $F_L$ be the class of all $L$-Lipschitz functions on $[0,1]$. Let $D$ be a joint distribution on $[0,1]\times\mathbb{R}$, from which we sample $n$ iid copies $(X_i,Y_i)$. Given any $...
3
votes
0answers
166 views

How useful is program search in the field of programming-language theory?

I've been thinking: computing systems such as the Lambda Calculus and its variations are usually very simple and can be implemented in as few as ~80 lines of Haskell code. There is a self-interpreter ...
3
votes
0answers
69 views

Lower Bound on Zero-order Regret

Here is a brief summary of the experts framework: Given $n$ experts who either give correct or wrong advice for each round $t\in [T]$, an algorithm is required to give a best prediction for each round ...

1 2
3
4 5
7