Questions tagged [neural-network]
Artificial neural networks (ANN), are composed of 'neurons' - programming constructs that mimic the properties of biological neurons. A set of weighted connections between the neurons allows information to propagate through the network to solve artificial intelligence problems without the network designer having had a model of a real system.
4,338 questions
2 votes
0 answers
15 views
Combining (interrupted) time-series analysis and machine learning to predict how an intervention will perform
I am working on a project that requires machine learning analysis, but I'm new to the field and still learning about different models. I just wanted to ask the community about the best model or ...
- 21
4 votes
1 answer
35 views
Possible to Improve Reconstruction Quality and Accuracy with VAE?
I am training a VAE architecture on microscopy images. Dataset of 1000 training images, 253 testing images. Images are resized to 128x128 input or 256x256 input from original resolution which is ...
- 41
0 votes
0 answers
58 views
Why bias value is critical to successful learning?
Given basic elements of a neuron(as below) with a bias value: I learnt that, a bias value allows you to shift the activation function(say sigmoid function) to the left or right, which may be critical ...
- 199
1 vote
0 answers
14 views
Scalar versus 2-element output for binary classification models
When building a binary classification model using a neural network, you have two options for outputs: output a single number from 0 to 1 using sigmoid activation, or output a probability distribution ...
- 31
1 vote
0 answers
13 views
How to interpret an unstable learning curve on a model tuned with Hyberband Tuning?
I have used Hyperband automatic tuning for an ANN model to predict price. After running the model with the automatic tuning, I am obtaining an R2 score of 1.00 that suggests overfitting, however, I am ...
- 11
2 votes
0 answers
26 views
What is the state-of-the-art of learning on streaming and non-stationary data?
I am trying to understand what are some of the methods to learn on streaming non-stationary data? Some examples of such data include Fraud Detection, Salary Prediction from browsing habits etc. The ...
- 31
-2 votes
1 answer
106 views
how to tell whether artificial neural network is suitable for a particular problem?
Usecase - Gmail detecting spam format emails is a simple NLP system based on some clear rules. Doesn't rely on artifical neural network. No training involved. Usecase - Gmail predicting auto-...
- 199
4 votes
1 answer
632 views
Why does the function of a neuron need to be in range [-1,1] or [0,1]?
The concept of a function includes three major components: a set called “domain”, a set called "codomain" and a rule which for each element of a domain points to (puts into a correspondence)...
- 199
6 votes
4 answers
158 views
Why large value gradient slow down training?
Why do we generally use activation functions with only limited range in neural networks? for e.g. $sigmoid$ activation function has range $[0, 1]$ $tanh$ activation function has range $[-1, 1]$ ...
- 199
6 votes
1 answer
85 views
Why normalizing outputs hurts the training?
I was wondering why normalizing outputs hurt the model training? In my case I have my outputs that are between 0 to 1 when I normalized them with MinMax() or StandardScaler() training basically dies ...
- 345
1 vote
0 answers
35 views
Is it really necessary to enforce constraints on ouputs in neural network?
I have a question. I'm doing a regression and I have 20 outputs where their sum is equal to 1 and also they are non-negative. I thought since their sum is equal to 1 maybe I can predict first 19 ...
- 345
5 votes
2 answers
261 views
How can I identify which activation function is the best for my neural network based on inputs and outputs?
My professor said I shouldn't use blind sense in neural network and I should choose activation functions carefully based on my inputs and outputs and their constraints. In the project I have the ...
- 345
6 votes
1 answer
147 views
How to implement this simple recommender system in Keras?
I've been working in data science for a long time, but very rarely have I been called upon to implement an ML algorithm; I've just ran other people's libraries. I'm trying to pick up the skill. I'm ...
- 305
0 votes
1 answer
58 views
when testing with shuffled data, accuracy is high, but when testing with unshuffled data, accuracy is low
To be clear, I shuffled my data when I trained it. It is only the testing data that I modified to be unshuffled, and found that accuracy tanks. (i also used the same data for training and for testing)
- 71
7 votes
2 answers
221 views
Combining image and tabular data for a binary classification task
I'm working on a binary classification task where the goal is to determine whether a tissue contains malignant cells Each instance in my dataset consists of a microscope image of the cell a small set ...
- 331