Questions tagged [machine-learning]
Questions about computer algorithms that automatically discover patterns in data and make good decisions based on them.
1,178
questions
0
votes
1
answer
18
views
Reinforcement Learning Reward Function for Optimizing Golf Aim?
I read this article, mentioning that either here, or StackOverflow would be the best places to ask generic machine learning questions, however, if the question isn't programming specific with a ...
- 103
0
votes
0
answers
8
views
Can I minimize a mysterious function by running a gradient decent on her neural net approximations? [closed]
A cross post from on AI StackExchange, and MathOverflow
So I have this function let call her $F:[0,1]^n \rightarrow \mathbb{R}$ and say $10 \le n \le 100$. I want to find some $x_0 \in [0,1]^n$ such ...
- 101
-1
votes
0
answers
24
views
the sum of probability of words in a sentence to be 1 in ngram [closed]
why add <s> and </s> to the start and end of a sentence can make the sum of the probability of words in a sentence ...
- 1
0
votes
0
answers
10
views
What are the set of actions in reinforcement learning?
I'm reading https://ai.stanford.edu/~ang/papers/icml04-apprentice.pdf
In 2. Preliminaries, they claim that the reward function takes a state in S to an action
$$R : S \rightarrow A$$
But in the next ...
- 101
1
vote
1
answer
21
views
Good Practice : Which VAE latent variable to use in dimensionality reduction
I'm trying to use a VAE (CNN-VAE to be exact) to reduce the dimension of some images I have in a dataset.
I successfully trained my VAE, but now I'm not sure of the latent variable to return.
Should I ...
- 73
1
vote
0
answers
39
views
Regarding constant * opt approximation in agnostic learning
In standard agnostic learning, we assume that there is a concept class $C\subseteq \{c:\{0,1\}^n\rightarrow \{0,1\}\}$. Given samples from a distribution $D:\{0,1\}^n\times \{0,1\}\rightarrow [0,1]$, ...
- 11
1
vote
1
answer
31
views
How can I understand the proof of the VC dimension of half-spaces in d-dimensions?
Statement : A half space is set of all points on one side of a linear separator, i.e., a set of the form $\{x \mid w^{T}x \ge t\}$. The VC-dimension of half spaces in $d$-dimensions is at least $d+1$.
...
- 35
0
votes
0
answers
14
views
What exactly is meant by the term"model" in machine learning?
I am very confused. What is the meant by the term"model" in machine learning?Is it any equation/expression or any software/program or a vector?? I have tried to explore google but couldn't ...
- 129
0
votes
0
answers
21
views
Relation between machine learning and computer vision?
What is correct concept in regards to relation between machine learning and computer vision?
As shown red encircled in attached snap, computer vision is a subset of machine learning
or
Is it an ...
- 129
0
votes
0
answers
10
views
Editing facial features
Photoshop offers some great image processing filters under the term "Face-Aware Liquify".
I want to integrate such techniques into an automated processing system I am currently implementing (...
- 121
2
votes
1
answer
37
views
Why are restricted Boltzman machines not considered useful?
I've been under the impression that RBM can be used for a variety of applications and a cursory search on google seems to confirm this. But I've heard from multiple people (including a peer reviewer) ...
0
votes
0
answers
26
views
How to detect touching on a glass with Object Tracking?
I am working at a Biomedical Center right now and my task is to track both pawns of rats and count the number of times they touch the glass surface.
To elaborate. I have a setup that contains one ...
- 1
2
votes
1
answer
378
views
What are common ways to distinguish between local minima and long run-time in hyperparamteer optimization
I'm using Bayesian optimization for a pretty cost expensive function in the context of neural networks in order to optimize hyperparameters for the neural net.
Is there a general, quantitative way to ...
1
vote
1
answer
53
views
Efficient algorithm for finding linear combination of piecewise functions
While porting an algorithm, I having a bit of a problem with finding an efficient algorithm for finding a linear combination of piecewise functions. The procedure is described in the Table 3 of the ...
- 113
0
votes
0
answers
35
views
resolving in CDCL
When resolving in Conflict driven clause learning, it is the case that if you resolve a conflicting clause with the reason of the negation of one of his literals, then this results in another ...
- 55
0
votes
0
answers
11
views
Doubt about Subword Regularization algorithm using Unigram Language Model [Kudo, 2018]
I've gone through the paper
Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates
link: https://arxiv.org/pdf/1804.10959.pdf
The algorithm is as ...
1
vote
0
answers
14
views
Are these representations of the Bellman Equation equivalent?
I've found two slightly different Bellman Equations. Are they totally equivalent?
I see the one on the bottom has an s' in the reward. Does this or anything else about the groupings change anything ...
- 119
1
vote
1
answer
16
views
Is there a good rule of thumb for regularization constants?
I'm trying to train some linear logistic regression models and I need regularization. My models contain around 4000 features.
I know that without regularization, a good rule of thumb is to have 10x ...
- 111
-1
votes
1
answer
41
views
Where should I start for making an AI Bot that processes on screen prompts
I would love to learn visual AI and what better way than creating a bot to play offline games. The problem is with no experience in AI other than Azure through my college, I have no experience.
Does ...
0
votes
0
answers
19
views
On the bounds of number of iterations of generative adversarial networks
I've been wondering whether there is any analysis on the bounds of number of iterations of generative adversarial networks (GAN), where here 1 iteration refers to 1 step of generator + discriminator. ...
- 23
0
votes
0
answers
15
views
Why is LSTM hidden state called hidden?
If I undersand correctly, in a LSTM, the output is the concatenation of the consecutives hidden states
$\text{output} = concat_{1\leq ï\leq n}(h_i)$.
But then why are those hidden states $h_i$ called &...
- 126
0
votes
0
answers
17
views
Learner Algorithm Time & Sample Complexity
Let $X=R^{2}$. Let $u=\left(\frac{\sqrt{3}}{2},-\frac{1}{2}\right),\
w=\left(-\frac{\sqrt{3}}{2},-\frac{1}{2}\right),\ v=\left(0,1\right)$
and $C=H=\left\{h\left(r\right)=\left\{\left(x_{1},x_{2\ }\...
- 221
0
votes
1
answer
56
views
Cosine distance in a space, and cheating?
I read an example of using cosine distance in RGB space, and it pointed out that (eg.) dark red and light red have a cosine distance (CD) of zero because CD only gives you the angle between vectors ...
- 199
1
vote
0
answers
26
views
pseudo dimension of the minimum of functions
Suppose a real-valued function class $\mathcal{F}$ with pseudo dimension less than $d$, I am wondering what is the pseudo dimension of the following function class
\begin{equation}
\mathcal{F}_2 = \{\...
- 111
0
votes
0
answers
16
views
Sensitivity analysis of Multi-Layer Perceptron Neural Networks
I have used Multi-Layer Perceptron Neural Networks to do a binary classification. Now I want to perform a sensitivity analysis. I am planning to follow a similar approach to this paper.
In case the ...
- 1
0
votes
0
answers
7
views
Calculate the joint probability of two nodes in Markov Random Field
I have a Markov Random Field (MRF) model as described in this picture.
I know the distribution factorizing over the graph is given by $$ p(x_{1:N}) = \frac{1}{Z} \prod\limits_{n=2}^{N} \psi_{n-1, n}(...
0
votes
0
answers
111
views
How many model parameters do R-CNN, Fast R-CNN and Faster R-CNN have respectively?
I am making research on object detection and I would like to know the number of parameters these three models obtain.
- 51
1
vote
1
answer
25
views
K-Modes Clustering with Partially-Overlapping, Variable-Length Data?
I'm working on a project that's attempting to cluster books using machine learning. I'm using the K-Prototypes algorithm for clustering data that has both numerical and class-based data. Under the ...
- 113
0
votes
1
answer
13
views
How much faster would a semantic segmentation model be with just 2 classes compared to 100 classes?
Let's say I have a semantic segmentation model that distinguishes between 100 classes of objects, and the speed of running the model is 1 image per second.
Now let's say I take the same model ...
- 1
1
vote
1
answer
73
views
When to stop training a back propagation algorithm based on the MSE?
I have a fully connected neural network that consists of 3 inputs + bias, 4 neuron hidden layer, and 2 layer output, I am using the sigmoid activation function on the hidden layer only and use the ...
- 11
1
vote
0
answers
12
views
Blackwell-Rao applications to Machine Learning and AI
We learned recently in a graduate statistic course about Blackwell-Rao of an unbiased estimator statistic T; other methods for parameter estimation includes Maximum Likelihood and Method of Moments ...
- 111
1
vote
0
answers
14
views
Multimodal machine learning dataset fine tuning for binary classification problem
Hello i am new to machine learning, i am working on a multimodal dataset, basically memes dataset, The data consists of an image and an overlay text, i have no clue how to feed a model both of the ...
1
vote
1
answer
25
views
Lagrange Multipliers and Hard Margin SVMs
With hard margin support vector machines (SVMs), it suffices to find the critical points of the Lagrangian $L = \frac{1}{2}||\theta||^2 - \sum_{n=1}^{N} \alpha_n (y^{(n)} (\vec{\theta}^T\vec{x}^{(n)} +...
- 11
1
vote
1
answer
26
views
Which artificial neural network model can I use when trying to predict gene expression?
I need help choosing a model for my data. I have 50 genes that I need to test for their expression levels. I was thinking of implementing either an SVM or a multilayer perceptron model where the ...
- 13
0
votes
0
answers
33
views
How does TRANSFORMER DECODER WITH MEMORY-COMPRESSED ATTENTION work?
I came across TRANSFORMER DECODER WITH MEMORY-COMPRESSED ATTENTION (T-DMCA) model in the paper "GENERATING WIKIPEDIA BY SUMMARIZING LONG SEQUENCES" (https://arxiv.org/pdf/1801.10198.pdf). ...
0
votes
0
answers
19
views
Kitti dataset Pi matrix. Pi is the same with Pi_rect?
Do i understand right that Pi in KITTI calib.txt files is the same with Pi_rect in their paper?: Pi_rect from paper
But what i can't understand it why last column in projection matrix in the paper ...
- 1
0
votes
0
answers
25
views
Graph algorithm for propagating a feature
I have the following problem:
Given a graph of persons and their relationships.
Suppose I have a classification of the risk of some of those persons.
How can I calculate the classification of the ...
- 101
0
votes
1
answer
48
views
If all words have the same frequency, is the generated Huffman tree a balanced binary tree?
If all words have the same frequency, is the generated Huffman tree a balanced binary tree? At the same time, is the generated Huffman tree a complete binary tree?
- 11
0
votes
0
answers
22
views
What is the best way to predict a continuous score based on image characteristics(pixels)?
I have a dataset with 9912 images, what is the best way to train them based on the pixels and features of the images to be able to predict a continuous target ranging from 1 to 100? The goal is to ...
- 1
0
votes
0
answers
10
views
Definitive source on energy based models
I'm currently working on a thesis based masters on graph generation. For this I need to learn about a variety deep generative approaches, including GANs and autoencoders. One of the approaches I need ...
1
vote
0
answers
30
views
Precise definition of Universal Learner in Machine Learning
It is surprising to me that I cannot find a precise definition of universal learner on the internet. I can guess what it should bebut I don't want to make a mistake, therefore I have come here.
Here's ...
- 11
0
votes
0
answers
34
views
Dependency analysis in a neural network
I was reading [1] about verifying neural networks. $n_{i,r}$ refers to the rth neuron in layer i and all neurons have the ReLU activation function. A neuron $n_{i,r}$ is said to be strictly active if ...
- 13
0
votes
0
answers
8
views
How a model can handle removed nighttime data in future predictions?
I want to use machine learning models for predicting one day ahead output power. In this model, I would like to remove all nighttime data because they do not make sense in prediction, but the problem ...
- 41
1
vote
1
answer
73
views
Labeling an image mask as data for object detection?
I am new to machine learning but had a question about a labeling method. If I had the following two images:
Is there a way to use the second image as the label for the first one (i.e. anything in ...
1
vote
1
answer
26
views
Linear encoding of a feed forward neural network
I was reading [1] about reachability analysis of a feed forward neural network (FFNN). The paper encodes a FFNN as a linear programming problem. Suppose $x^{(i)}$ is the vector output of the ith layer,...
- 13
0
votes
0
answers
37
views
Out-sample error for disjoint input spaces
I am trying to understand the fundamentals of Bayesian machine learning and I found an interesting task, which I have tried to solve:
We are in the setting of a binary classification problem with $Y=\{...
0
votes
0
answers
19
views
Different results in standard deviation of test set and cross validation
I have used KNN for predicting the output power of solar panels. The results are shown below:
Test Set:
MAE = 0.044
MSE = 0.004
Standard Deviation = 0.272
...
- 41
1
vote
1
answer
62
views
MAE greater than MSE - is it possible?
I have used KNN for predicting the output power of solar panels. I have done feature scaling based on normalization for all data points. The problem is that the MAE is equal to 0.08, and MSE is equal ...
- 41
1
vote
1
answer
16
views
Is it really impossible to use K-fold cross validation for time series data?
I would like to know if it is impossible to use K-fold cross-validation for time series of data. It has been discussed that it is not reasonable to predict past from future, but I think this does not ...
- 41
2
votes
1
answer
20
views
Clarification about RNN encoder-decoder equation
In the paper by Cho et.al., section 2.3 details the equations for the modified LSTM cell in RNN used in the paper's implementation. The equation in question is :
Here, the output of the reset gate (r)...