Questions tagged [machine-learning]
Questions about computer algorithms that automatically discover patterns in data and make good decisions based on them.
0
votes
0answers
10 views
Can the test set of attributes be a subset of the training set's attributes?
I'm currently writing an application for a trick-based card game, where agents are assigned points based on the accuracy of their predictions of how many hands they're going to win. The number of ...
0
votes
0answers
8 views
What is the practical limit to how many object classes you can detect with Faster RCNN?
I am trying to follow this tutorial where the Faster-RCNN-Inception-V2-COCO model from TensorFlow's model zoo is used to detect playing cards. I was wondering what is the practical limit to the number ...
1
vote
1answer
23 views
what does this phrase mean: “train a policy network”
I am familiar with the basics (and perhaps a substantial amount of basics) of imitation learning and reinforcement learning. In IL (imitation), we take demonstrations from an assumed expert, which we ...
7
votes
1answer
56 views
Why do mainstream speech models no longer require a personalized training step?
Back in the Windows XP era, when setting up Windows OS-built-in speech/dictation, I had to speak out a bunch of programmed-in text samples to the speech-to-text engine to personalize my voice profile.
...
1
vote
0answers
14 views
How to add important events as input in neural network?
I'm quite new to neural networks so I apologize if this question is too basic/doesn't really make sense.
I have a financial time series dataset and I have a binary variable which is 0 if no important ...
0
votes
0answers
19 views
Batch sizing for convolution neural networks — powers of 2 or powers of primes?
The conventional wisdom for convolution neural networks (CNNs) is to make the batch size a power of 2 because of hardware utilization/optimizations done in the convolution layers.
A similar logic ...
1
vote
0answers
13 views
Is a linear classifier convex?
Is the optimization of a linear classifier convex?
Is there any local optima or saddle points for a linear classifier?
0
votes
0answers
11 views
How does neural net complexity relate to other complexity measures?
In neural networks, "weight regularization" is often used as a so called "complexity penalty" in order to make sure that the network generalizes better from training data.
Similarly, in "program ...
0
votes
0answers
32 views
What is the relationship between pairwise loss and centroid loss?
What is the relationship between pairwise loss and centroid loss? Under what conditions you would expect them to give similar behavior? Under what conditions would they give different behavior?
1
vote
0answers
14 views
How to compute the sample update error?
In the book Reinforcement Learning An Introduction，Chapter 8.5，there is an example that compares the efficiency of expected and sample updates:
According to the author, "In this case, sample updates ...
0
votes
0answers
6 views
OR functions and SQ-Learning
Anyone can describe or give a reference which has a clear description and detailed proof of the SQ-Learning algorithm for the OR class of Boolean functions?
0
votes
0answers
14 views
How should I represent an unordered set of dense vectors as input to a neural network?
My dataset consists of 10000 (x,y) pairs where x is an unordered set of 27-dimensional vectors with a total of 8 elements, and y is a different unordered set of 27-dimensional vectors with a total of ...
-1
votes
0answers
18 views
Understanding the computational power of neural networks
It is known that a recurrent neural network with rational weights is computationally equivalent to a Turing Machine (a proof can be found in this paper).
I don't understand how is it possible, it ...
0
votes
0answers
8 views
A question on Relative Entropy Inverse Reinforcement Learning
I am reading Boularias' Relative Entropy Inverse Reinforcment learning (http://proceedings.mlr.press/v15/boularias11a/boularias11a.pdf) and I do not understand a paragraph of the premise of the method ...
0
votes
1answer
24 views
How can I represent the following mathematically? (Eigenfaces reconstrcution)
I got the following from a web site:
The feature vectors represent each image as a linear combination of
the eigenfaces defined by the coefficients in the feature vector; if
we multiply each ...
2
votes
0answers
37 views
Card dealing problem with constraints (blacklisting),
A friend of mine and I are trying to teach a bot play a card game (bela)
We are using monte carlo tree search (MCTS) to estimate the probability of winning hand in regards to multiple possible (!...
1
vote
1answer
72 views
(OCR ) How to Recognise Handwritten fractional numbers using Neural networks
I want to be able to recognise handwritten math numbers using images of the numbers , i was able to do create a ANN model for recognising simple decimal numbers , but i have no idea on how to ...
1
vote
1answer
85 views
Help needed for understanding proof of No Regret Multi Armed Bandit Algorithm
I was reading Elad Hazan's book on Online Convex Optimization(http://ocobook.cs.princeton.edu/OCObook.pdf) and am facing difficulty understanding the proof given for the No regret algorithm for MAB (...
-1
votes
0answers
26 views
Explanation required about these Ghostly Images
I am reading the research paper “Eigen faces for Recognition”. Paper Link. In Figure 2, paper shows the seven Eigen faces having white and black spots on them. Is there any significance of these white ...
1
vote
1answer
43 views
Do data points mean Eigenfaces in higher dimensional space?
I saw the following animation at making sense of PCA
, which shows blue data points.
I am reading a paper on Eigenfaces which says that:
"a typical image of size 256 by 256 becomes a vector of ...
1
vote
0answers
16 views
Machine learning for VLSI floorplanning tool
I have an educational assignment to make an VLSI floor-planning tool.
Can I use machine learning in some part of the algorithm? For example, I was reading the book Algorithms for VLSI Physical Design ...
0
votes
1answer
78 views
What is the difference in SMO algorithm for SVM and SMO for one class?
Please let know if this is not the correct forum to ask this question. If not can anyone please tell where can I ask this question?
I am trying to understand the difference between the paper :
https:/...
1
vote
1answer
30 views
Problem with Understanding “correlated attributes into a set of values of uncorrelated attributes” in PCA
I am studying PCA. I have a problem in understanding the following concept:
What is meant by transforming "correlated attributes into a set of values of uncorrelated attributes" in Principal ...
1
vote
0answers
18 views
Is there an equivalent of “control flow” in deep learning systems?
Traditional programs work significantly through control flow. I know that there are simple algorithms that have no control flow at all (i.e. they just perform a fixed sequence of executions), but I ...
0
votes
0answers
4 views
Looking for a specific type of ADMM iterates
For a $k-$dimensional optimization variable $b \in \mathbb{R}^k$ say the objective is given as, $$f(b) = \langle b, v \rangle + \langle b , Ab \rangle + \lambda \Vert b \Vert_1$$
for some parameter ...
0
votes
0answers
12 views
Is the optimization of the Gaussian VAE well-posed?
In a Variational Autoencoder (VAE), given some data $x$ and latent variables $t$ with prior distribution $p(t) = \mathcal{N}(t | 0, I)$, the encoder aims to learn a distribution $q_{\phi}(t)$ that ...
1
vote
1answer
46 views
Show that the Laplace smoothing for bigrams is a valid probability distribution
If we consider any smoothing technique like laplace or delta smoothing. Intuitively we can see that the we are stealing from sequences with non zero probablity and re distribute to sequences with zero ...
1
vote
0answers
30 views
Proof of perceptron convergence theorem for ZERO threshold?
The generalized perceptron convergence theorem is for a defined threshold T.
When you do the maths it all comes to an upper bound and a lower bound.
The lower bound looks like this!
Therefore
...
1
vote
1answer
21 views
Checking if a kernel is valid
The kernel is $K(x,z) = \sum_{i=1}^D (x_i+z_i)$
My approach was trying to express $K = \phi(x)^T\phi(z) = (x_1 x_2 ... x_D \quad 1 1 ...1)(1 1 ...1\quad z_1 z_2 ... z_D )^T$ where $\phi$ is 2Dx1 ...
2
votes
1answer
36 views
Is Artificial General Intelligence possible with our current machine learning models? [closed]
In other words, is artificial human level intelligence not possible yet just because of limitations in processing power and amount of data required to train the models? Or we don't have the knowledge ...
0
votes
0answers
14 views
how many hidden layers in an n-layer neural network
Simple terminology question: (should be easy to answer)
How many hidden layers does an n-layer neural network have?
I believe the answer is n-1. For example a single layer perceptron has no hidden ...
0
votes
0answers
14 views
Applications of signed permutations to machine learning
Are there some applications of signed permutations to machine learning? I searched on google and only found one paper. Thank you very much.
1
vote
1answer
30 views
Matching/finding mathematical plot images
I'm trying to come up with a method to match a given mathematical plot against a database of other plots. To make it more specific: plots are generated in R as PNG and might have different dimensions. ...
0
votes
0answers
15 views
momentum term in back propagation
back propagation with momentum term means we update the weights like so:
$\Delta w_{i,j}(n)= \eta*\delta_j(n)*y_i(n)+\alpha*\Delta w_{i,j}(n-1) $
what do we intialize $\Delta w_{i,j}(0)$ to?
3
votes
1answer
24 views
Why doesn't this derivation of the margin in a SVM give the correct result?
I'm trying to derive the optimization objective for an SVM (namely $1/\|w\|$), but I'm running into a little trouble. I've already read this question, which has certainly offered a lot of insight into ...
1
vote
1answer
19 views
The notion of PAC in approximation algorithms
In computational machine learning, the notion of Probably Approximately Correct means that (generally speaking) we can find (or "learn") with a high probability a function which has "low error".
Is ...
0
votes
1answer
17 views
Choose the best classifier to predict the label of strings of a regular language
I have to tackle this problem:
I have some strings that are my training set. These strings belong to a regular language corresponding to a deterministic finite automata (hidden namely I don't now it, ...
0
votes
0answers
28 views
FastText how it works?
So my question is how Supervised FastText works for the most part. I understood in the original paper they use bag of n-grams for features, but then they released a paper with enriching the word ...
2
votes
1answer
126 views
Mathematical proof for why gradient descent algorithm always converges
I am currently learning machine learning and I stumbled across gradient descent. I understand why the algorithm always converges to the global/local minimum when the learning rate is small enough in ...
1
vote
0answers
24 views
What are good examples of computational theories for A.I. according to David Marr's Definition?
I was reading David Marr's "Artificial Intelligence-A Personal View" and he talks about "computational theory of AI" or what he laters labels as "Type 1" Theory. He provides the example of Chomsky's ...
-3
votes
2answers
163 views
Does deep learning infer P = NP?
The question comes from the following scenario,
assume we have the traveler problem which is NP (the one where a traveler wants to visit all countries with the lowest cost(by summing up all flights))
...
1
vote
2answers
123 views
What is the “spatial information” in convolutional neural network
deep learning research papers always claim that deeper layers of CNN have good "semantic information" but poor "spatial information". What is the spatial information exactly. Is that some activations ...
3
votes
0answers
26 views
How to compute the loss and backprop of word2vec skip-gram using hierarchical softmax?
So we are calculating the loss
$$J(\theta) = -\frac{1}{T}\sum_{t=1}^T\sum_{-m \leq j \leq m} \log P(w_{t+j}|w_t;\theta)$$
and to do this we need to calculate
$$P(o|c) = \frac{\exp(u_o^Tv_c)}{\sum \...
1
vote
1answer
99 views
VC dimension of finite unions of one-sided intervals
What is the VC dimension of $k$ finite unions of one-sided intervals:
If we take 3 one-sided intervals like $(-\infty, a_1] $, $(-\infty, a_2] $ and $(-\infty, a_3] $, I think union of these ...
1
vote
1answer
72 views
Cost of computational representation in PAC-learning definition
I'm currently reading Foundations of Machine Learning by M. Mohri, A. Rostamizadeh, A. Talwalkar and according to their definition a concept class $C$ is said to be PAC-learnable if
$$Pr_{S \sim D^m}[...
1
vote
0answers
24 views
How to get old top news? [closed]
I am not sure this is the right forum to ask.
For a machine learning training, I need a dataset of old top news by keywords that can be organized by date. From where I can download it? I have seen ...
3
votes
1answer
46 views
(DROP) Data Reduction Algorithm - How it works?
I am studing a PHD framework which the propose is to reduce the dataset with the most representative samples for training a classifier. Maybe I am loosing something, but I could not undestand a ...
3
votes
1answer
34 views
Weird behaviour of softmax derivative?
I have been implementing some neural networks in MATLAB and recently I noticed a weird thing while implementing softmax derivative:
Setting the derivative to one, rather than using the actual ...
0
votes
1answer
25 views
How to calculate the weight between neurons in ANN?
I am currently learning Supervised ANN training using Backpropogation and I am stuck in this exercise. I calculated the δA using the equation at the bottom of the screenshot, however, I am unable to ...
5
votes
2answers
92 views
Why do AlphaGo and AlphaGo Zero include board history in the input features
Both AlphaGo and AlphaGo Zero include prior board states as input features (the "Turns Since" planes for AlphaGo, and the repeated 8-step history planes for AlphaGo Zero).
What is the purpose of ...
