Search Results
| Search type | Search syntax |
|---|---|
| Tags | [tag] |
| Exact | "words here" |
| Author |
user:1234 user:me (yours) |
| Score |
score:3 (3+) score:0 (none) |
| Answers |
answers:3 (3+) answers:0 (none) isaccepted:yes hasaccepted:no inquestion:1234 |
| Views | views:250 |
| Code | code:"if (foo != bar)" |
| Sections |
title:apples body:"apples oranges" |
| URL | url:"*.example.com" |
| Saves | in:saves |
| Status |
closed:yes duplicate:no migrated:no wiki:no |
| Types |
is:question is:answer |
| Exclude |
-[tag] -apples |
| For more details on advanced search visit our help page | |
Results tagged with machine-learning
Search options not deleted
user 26948
Machine learning algorithms build a model of the training data. The term "machine learning" is vaguely defined; it includes what is also called statistical learning, reinforcement learning, unsupervised learning, etc. ALWAYS ADD A MORE SPECIFIC TAG.
1
vote
Accepted
How can I derive lambda and alpha in SELU activation?
I don't claim to understand the full proof provided in the paper. However, I do know that $\alpha$ and $\lambda$ are constants, not learned.
I thought self-normalizing meant the function would adj …
- 26.5k
2
votes
Accepted
Are models using satellite image inputs well-posed?
You could imagine a "sliding view function" $f(I, x) = V$ of some large input image $I$ and position $x$, and returns a cropped view $V$ of the image. This function could be implemented with linear in …
- 26.5k
1
vote
What is the name of an algorithm that does not use training data?
I think I've seen "non-learned" or "not learning-based" used at least a few times. "Hand-crafted" or "hand-engineered" or "hard-coded" are also in use. In some contexts, talking about a "naive" agent …
- 26.5k
6
votes
In what type of application a small training set would be important?
Generally speaking, the more data is available, the better. In fact, papers have found that training a model on 300 million images is better than only 30 million -- it's not the case that our models c …
- 26.5k
2
votes
Accepted
How to stabilize training in multi-objective reinforcement learning?
If the reward from each objective $1... k$ is $r_1, r_2, ... r_k$ set the goal to be maximization of $R = \min_i r_i$ instead of $R = \sum_i r_i$. By maximizing the minimum reward across all goals, th …
- 26.5k
1
vote
Accepted
Accuracy of K Nearest Neighbor in near-ties
It's true that KNN, when run on bogus data, can appear to find patterns which don't exist. However, you could say the same for soft-margin SVM, or logistic regression, or nearly any classification alg …
- 26.5k
3
votes
Accepted
Combining image and scalar inputs into a neural network
There are many ways to combine scalar and image inputs. In this particular paper, a diagram on the top of page 5 should explain everything. At some point in the convolutional network there are 64 feat …
- 26.5k
1
vote
Accepted
Hidden layers neural networks
Since you only have what looks like 5 input units for each network and what looks like one or two outputs, this is probably something which can be done with at most one hidden layer with 10 units or l …
- 26.5k
2
votes
Accepted
When training a neural network, how much does it help to sub divide the training output values?
Yes.
Providing fine-grained labels can help, and probably doesn't hurt, although it's hard to say exactly how much it would help.
To provide a concrete example, suppose most vehicles tend to have whe …
- 26.5k
1
vote
Accepted
Bolzman machine - sampling
In the positive phase, the gradient is $E[hx^T|x] = E[h|x]x^T$. Since $E[h|x]$ can be computed without sampling, there is no sampling involved in the first phase.
In the negative phase, the gradient …
- 26.5k
1
vote
Accepted
Label importance scale - Supervised learning
Depending on the optimization algorithm used, scaling can be an issue. With stochastic gradient descent, using a weight of 10/20/90 versus 0.1/0.2/0.9 is effectively the same as multiplying/dividing t …
- 26.5k
1
vote
in a CNN, why isn't the "type" information obtained at one layer lost by the next layer?
For classifying an object or animal, it's not necessary to remember the activations of the low-level neurons, which mostly learn to detect edges, corners, and gradients. So there's no problem with thr …
- 26.5k
7
votes
Accepted
Simplest way for ANN to learn F = MA?
Yes, the network will probably approximate some sort of multiplication, but it is unlikely to generalize outside the range of inputs you train it on.
You may have more luck learning and generalizing …
- 26.5k
2
votes
Accepted
Computing the posterior probability in VAE's decoder
For continuous $X$:
$X|z \sim \mathcal{N}(\mu = f(z), \sigma^2)$ where $f$ is the decoder of the network, and $\sigma$ is a hyperparameter of the model.
For discrete $X$:
$X|z \sim \text{Bernoulli} …
- 26.5k
2
votes
Accepted
Create a learning system that can generate highly structured data such as SVG graphics or HT...
Yes. This is known as structured prediction, and usually done with some sort of recursive neural net. Often for code, there is some inherent tree-structure which can be exploited if your RNN is able t …
- 26.5k