10 votes

Difference between machine learning and artificial intelligence

The subject areas Artifical Intelligence and Machine Learning (plus Data Science) are loosely defined, such that it is hard to make strict statements about how they relate. In the general case, it ...
Neil Slater's user avatar
  • 28.9k
9 votes

Difference between machine learning and artificial intelligence

Machine learning in layman terms is an algorithm that allows machines to identify patterns in data and then develop a model which can be used to predict unseen data. Artificial Intelligence is the ...
Ajay Sant's user avatar
  • 276
6 votes

What is Inductive bias?

This is a tentative general answer. Bias in statistics (and machine learning) is the difference between an estimated value and the true value. ie $\text{bias} = \hat{y}_{\text{estimated}} - y_{\text{...
Nikos M.'s user avatar
  • 2,333
5 votes

Difference between machine learning and artificial intelligence

Deep Learning is a subset of Machine Learning which is a subset of Artificial Intelligence. Machine learning is a particular approach for AI but not the only one. Symbolic Logic, Bayersian Statistics ...
Dhruv Mahajan's user avatar
3 votes
Accepted

A trick used in Rademacher complexity related Theorem

This requires hell of a derivation, but I liked the question :) My question is, why can we swap $z_i$ and $z'_i$? The key insight is that notation $S \sim \mathcal{D}^m$ is equivalent to $Z_1 \...
Esmailian's user avatar
  • 9,312
3 votes
Accepted

What does it mean when someone says "Most of the data science algorithms are optimization problems"

Most algorithms try to minimizes some objecive functions. For example, in linear regresssion, given $(x_i, y_i)$, we try to find $\hat{y}_i= \alpha_0 + \sum_{j=1}^d \alpha_j x_{i,j}$ and we we want it ...
Siong Thye Goh's user avatar
3 votes
Accepted

Deriving backpropagation equations "natively" in tensor form

Notation matters! The problem starts from: Given $\nabla a_j^{(k+1)} = \frac{\partial E}{\partial a_j^{(k+1)}}$ I don't like your notation! it's wrong in fact, in standard mathematical notation. The ...
Ehsan M. Kermani's user avatar
3 votes

Difference between machine learning and artificial intelligence

ML, by Tom M. Mitchell: A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, ...
3 votes

Difference between machine learning and artificial intelligence

A good example of AI, but not machine learning is evolutionary computation. Here instead of learning from experience (as in Tom M. Mitchell's definition) we have genotype changing in each generation ...
DmytroSytro's user avatar
3 votes
Accepted

Which neural network is better?

When you train a neural network, you usually use 3 sets: one for training, one for development, one for testing. Your training set is here for (obviously) training your model: the performance of your ...
Clef.'s user avatar
  • 146
3 votes

Where does AI/ML theories come into play when nowadays the AI libraries already so powerful?

First "linear algebra, statistics, complicated optimization" are not ML theories, they are mathematics toolkits that guides specific ML algorithms designs and improvement at operation level. ...
imadcat's user avatar
  • 276
2 votes

Difference between machine learning and artificial intelligence

As the great Tom Mitchell has said in his book "Machine Learning is the ability to learn without being explicitly programmed." Machine learning algorithms are widely employed and are encountered ...
surya rahul's user avatar
2 votes

What would be the best way to impute data?

You can treat the missing feature as the target variable of a sub-problem and create a classifier (e.g., a linear model, SVM, etc) for it.
Ryan Zotti's user avatar
  • 4,149
2 votes

Situations where advanced theoretical knowledge of ML helped solve a real world problem?

At the end of the day, as an applied data scientist you have a bag of tools you can use to solve business problems. It's helpful to know what your tools are capable of, where each is useful, and also ...
tom's user avatar
  • 2,248
2 votes

Difference between machine learning and artificial intelligence

Let's take the total Turing test as an example. A computer is often said to be intelligent if it can pass the total Turing test. A computer passes the test if a human interrogator, after posing some ...
Bayequentist's user avatar
2 votes
Accepted

Is it possible for a neural net to score as high as a different form of supervised learning?

Generally speak, no. Deep learning models struggle to compete when it comes to tabular data. If we head over to kaggle were people compete to build the best model we find that usually the best ...
Simon Larsson's user avatar
2 votes
Accepted

Multiple solutions with same minima in MLP with same weights

If one permutes the connections of the hidden layer ($d!$ ways to do that), and move and rename connections appropriately, then one effectively has the same MLP with the exact same minima, yet the ...
Nikos M.'s user avatar
  • 2,333
2 votes

How to input a list into my model and not have it care about order

If the range of possible integers is small, encode the presence of each integer as a boolean column in a feature vector. Example with a value range of 0-5. ...
Bert Kellerman's user avatar
2 votes
Accepted

Given M binary variables and R samples, what is the maximum number of leaves in a decision tree?

The maximum possible combinations with M binary variable is $$ 2^M $$ so essentially if all these values have different classes, then the number of leaves should be equal to $$ if \ R<2^M => R \\...
Harsh Sharma's user avatar
2 votes

Сan we say that the regression problem is essentially a classification problem with an infinite number of classes?

From a somewhat technical perspective, this is roughly correct: you can approximately solve a regression problem by solving the classification problem that you mentioned, but ... ... there is a ...
BanDoP's user avatar
  • 211
1 vote
Accepted

An ambiguity in SVM equations about misclassified data

Good point! Interesting consequence! Problem is the $a_n=0$ assumption, i.e. assuming misclassified points are not support vectors. Here is the flow. Slack variable $\xi_n$ is defined as $$\xi_n := ...
Esmailian's user avatar
  • 9,312
1 vote

A question on realizable sample complexity

We want to prove: If H is PAC learnable, then $\forall \epsilon, \exists C, \forall m \geq m_2:=Clog(1/\epsilon)(m_1+1/\epsilon^2), E[L] \leq \epsilon \mbox{ (a)}$ where $m_1:=m(\epsilon/2,1/2)$ ...
Esmailian's user avatar
  • 9,312
1 vote

Decentralized machine learning

Using federated learning the model training does not require the whole data to be present at a centralized server instead the model training is decentralized such that the model gets trained ...
Amit Rastogi's user avatar
1 vote

Theoretical basis for neural network "effort"

You asked several questions, I'll answer the one about which specific distributions are easier to learn. Information Theory would predict that $[0,0,0,1]$ would far easier to learn than $[\frac{1}{2},...
Brian Spiering's user avatar
1 vote
Accepted

What is the opposite of baseline?

The term is Oracle. Some references: SO question describing the term Scientific articles related to machine learning using the term
noe's user avatar
  • 26.5k
1 vote

Can a neural network be of variable depth?

"Wide and Deep network" was used in a 2016 paper. $\hspace{5cm}$Wide & Deep Learning for Recommender Systems You may create using keras Concatenate layer ...
10xAI's user avatar
  • 5,584
1 vote
Accepted

What is an object detection problem with only one class called?

If you are talking about "two-stage" obejct detectors like Faster R-CNN, note that the second phase is not only for classification, but to obtain more accurate results (https://stackoverflow.com/a/...
Graph4Me Consultant's user avatar
1 vote
Accepted

Gradient for starting Backpropagation

This is because of the loss assumption is the (Mean) Squared Error $\mathcal{L} = (\hat{y} - y)^2$ and the derivative is $$ \frac{\partial}{\partial \hat{y}} \mathcal{L} = 2 (\hat{y} - y) $$ which ...
psiyumm's user avatar
  • 141
1 vote

Is the search space of Hyperparameters Continuous or Discrete?

Continuous means only you have continuous variables. It can be convex or concave. It might not even be differentiable. Gradient descent only applies to differentiable convex problems (or convex ...
Piotr Rarus's user avatar
1 vote
Accepted

What is the name of this statistical interaction?

You might want to look at conditional entropy, H(A|B) and H(B|A).
MrMulliner's user avatar

Only top scored, non community-wiki answers of a minimum length are eligible