Stack Exchange Network

Stack Exchange network consists of 174 Q&A communities including Stack Overflow, the largest, most trusted online community for developers to learn, share their knowledge, and build their careers.

Visit Stack Exchange

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 ...

0
votes
0answers
14 views

How does the loss function for semantic segmentation networks (like FCN) work?

Just want to understand how the cost function works when performing semantic segmentation. I know that for simple classification networks, the output is a fully connected layer equal to the number of ...
0
votes
0answers
9 views

My TD-backprop algorithm doesn't work

In the previous discussion I have tried to solve the TTT game with Q-learning with tables. Now I have tried to use Neural Network like function approximator and following these articles (for game of ...
1
vote
1answer
19 views

Neural network model for sparse multi-class classifier on Tensorflow

The problem I'm trying to solve is the following: the data is Movielens with N_users=6041 and N_movies=3953, ~1 million ratings. For each user, a vector of size N_movies is defined, and the values ...
2
votes
1answer
24 views

How can we use Neural Networks for Decision Making intead of Bayesian networks or Desicion Trees?

I am working on Decision Making in Self driving cars and I am wondering how I can use Neural networks (is there any type) ? that can repleace or mimic the bayesian networks or Decision Tree for ...
0
votes
0answers
16 views

Using both positive and negative values as neural network input?

In neural networks, we sometimes convert the input to z-scores. However, z-scores contain both negative and positive values, if we use such numbers as input, it seems that in some cases the neural ...
0
votes
1answer
15 views

How can I perform backpropagation directly in matrix form?

I had made a neural network library a few months ago, and I wasn't too familiar with matrices. So, instead of performing matrix dot products (between weights and inputs, then adding a bias matrix), I ...
0
votes
1answer
13 views

Scaling features in artificial neural networks

So it is a well known thing that it is a good idea to scale features/training samples in the training set, so that the values do not differ too much in the absolute sense. For example we want to train ...
0
votes
1answer
18 views

Sequence models word2vec

I am working on data-set with more than 100,000 records. This is how the data looks like: ...
2
votes
4answers
39 views

Why is eulers number used as a constant in sigmoid

I was asking myself why eulers number was used in the sigmoid function 1/(1+e^-x) instead of any other constant like for example ...
1
vote
1answer
58 views

Model Not Learning with Sparse Dataset (LSTM with Keras)

This classification problem is apparently simple and I have no idea why it's not working, perhaps I'm doing a conceptual mistake. I'm trying to make a predictor which will classify minutes on a clock ...
0
votes
1answer
12 views

Using neural networks with jumps in stock returns

I am using an LSTM network to analyse stock return patterns. A problem is that, there is usually huge jumps in stock returns but if you are only using the trading data, the jumps would seem pretty ...
-1
votes
0answers
18 views

Simple ANN in C++ for newbies [on hold]

I am working on a project. I need a simple ANN in C++ which has around 90% accuracy. It could be on any dataset. I have looked up the internet but it seems like most of the resources are centered ...
0
votes
0answers
24 views

Do capsule networks have to be trained on different poses of an entity for them to work?

I have read about capsule networks and have failed to understand the following. A capsule network can identify objects at different poses(affine transforms) via its instantiation parameters.But my ...
0
votes
0answers
12 views

How to create autoregression model for multiple inputs and one output with time interval 120min using python

Here I have data with time import from csv file. I have three inputs (x,x1,x2) with actual output value (y) with time period. I want to predict the next value using past values with time using ARIMA ...
1
vote
1answer
20 views

How to implement keras LSTM time series [on hold]

I am learning how to implement Keras LSTM on a simple time series data. The dataset I'm using has $12$ columns and $300k$ rows. Each group of $200$ rows represents ...
0
votes
0answers
12 views

LSTM worse for shorter time series horizon

I got a very strange counter-LSTM-hype result. I have a long time series of multiple covariates, and I train a vanilla LSTM with subsequences of length from 15 to 75. I use cross-validation to ...
0
votes
0answers
11 views

Is it possible to fine tune a model on the training dataset to extract features for the validation dataset?

I have 2 models to train. The first is VGGFace, which is already trained. I have to fine tune it using my dataset (labeled frames where labels describe emotions). then, I am going to extract features ...
2
votes
2answers
18 views

What is the relation between input into LSTM and number of cells?

I want to train an LSTM network for time-series predictions, and want to get to the bottom of LSTM's. In my understanding, the number of cells in a single LSTM layer can vary. However, since each cell ...
1
vote
1answer
12 views

Stability of value function approximation in policy gradients

In DQNs, function approximation of the Q-values is unstable for correlated updates. In policy gradients with a baseline, will the value function of the policy not be plagued by the same correlated ...
1
vote
3answers
47 views

How to find the most important attribute for each class

I have a dataset with 28 attributes and 7 class values. I want to know if its possible to find out the most important attribute(s) for deciding the class value, for each class. For example an answer ...
0
votes
0answers
12 views

Using GPU for genetic algorithm [Python] [on hold]

I'm training neural networks playing Snake through a genetic algorithm in Python, everything is fine but I noticed that it spends most of the time playing games in order to compute the fitness for ...
0
votes
0answers
31 views

How can I classify these aerial images?

I want to use R to classify high res aerial images (4 band tiffs). The images are of residential food gardens in Portland. I want to train a model to be able to identify if there is a food garden ...
0
votes
1answer
21 views

Can't understand Output shape of a Dense layer - keras

I am following few online tutorial to classify images and started off with dense layers as a starting point to classify cifar10 data. ...
0
votes
0answers
23 views

Train TicTacToe NN by human moves

Last few months I started with machine learnig little bit more than just reading articles, so I'm trying to understand it in practice. I started with tic tac toe game. It looks easy and funny, and it ...
0
votes
0answers
14 views

Understanding LSTM/RNN structure

In keras when we apply LSTM/RNN model, we specify the node [i.e.,LSTM(128)]. I have a doubt how it actually works. From the LSTM/RNN unfolding image or description, I found that each RNN cell take one ...
0
votes
0answers
14 views

Penalize Neural Network Common Output

To practice reinforcement learning, I have made a neural network class and have been trying to teach it to play a toy system of Pokemon. There are three types of pokemon, and each pokemon has access ...
1
vote
1answer
27 views

Neural network recommendations if only few features

Can there be some general recommendations for architecture of neural network if there are only a few features, say 2-5 features? What should be the number of hidden fully connected layers here? How ...
0
votes
0answers
14 views

Neural Network stimuli propagation

Is not clear to me how exactly the stimuli proagate through a neural network. It's pretty clear how it should work in a feed-forward network but not in a more complex one. If i have understood, if ...
3
votes
2answers
58 views

Depth of a Neural network

I was self-teaching myself. I totally understand why depth of a neural network affects the learning and how it differs than its width. But I am looking for some theoretical justification about it. ...
0
votes
1answer
32 views

Validation loss is lower than the training loss

I am using autoencoder for anomaly detection in warranty data. Architecture 1: The plot shows the training vs validation loss based on Architecture 1. As we see in the plot, validation loss is ...
2
votes
4answers
38 views

Is gradient descent slower for finite differences?

In gradient descent, we updated each parameter $\theta_i$ in the direction which minimizes a function $f(\theta_1,\theta_2,\dots,\theta_N)$ by doing $$\theta_1 \leftarrow \theta_1 - \alpha \frac{\...
0
votes
0answers
7 views

convolution neural network:representing vector in fully connected layer

please i would like to ask about representing feature vector in the fully connected layer in cnn. i have image and i cropped it into N segments and fed each one into cnn branch and get feature maps ...
0
votes
1answer
19 views

Why does my LSTM perform better when randomizing training subset vs. standard batch training?

I am training a simple LSTM network using Keras to predict time series values. It is a simple 2-layer LSTM. I get the best performance when I train on subsets of the training set that start at random ...
1
vote
1answer
15 views

How to build a classifier with a rejection class

Let's say I need to build a food classifier, and I want a rejection class for the inputs that are not food. What is the best way to do that? Should I just add a new class label that includes ...
0
votes
0answers
7 views

Support Vector Machines VS LSTMs: How well it is justifiable to use LSTM for its Generalization properties?

The question is pretty straightforward, How well one can justify using LSTMs(Neural Networks) for text classification task in terms of "Generalization" compared to classic support vector machines(SVM) ...
0
votes
0answers
8 views

Keras LSTM with wavelet transformed inputs grouping?

I am doing a project where I am using financial candle data (OHLCV). If I preprocess each column using Haar discrete wavelet transform, each float value is converted into 2 separate values, the ...
0
votes
0answers
18 views

Which loss and activation function at output layer is suitable for Multi target classification problem?

I am modelling a Multi-target classification problem which has 220 input features and 132 output features. Each output target has an integer value in between [0,1,2,3,4] .And for this I have applied ...
0
votes
0answers
13 views

Performance degradation from video compression

I have two datasets. The first are frames saved as pngs (lossless) from a live video feed, and the second are the same frames taken from an mp4 (H.264 compression). Training the same image classifier ...
0
votes
1answer
15 views

What type of neural network could emulate a binary to HTML conversion tool?

I've got a problem, which I thought could be solved by using a neural network: I've got a binary file and a tool that converts that file into a readable html file (probably a text file as well). How ...
0
votes
0answers
20 views

Unsupervised Pattern Mining

I have a relatively large transactional dataset ~1TB of data (around 30 columns). I am interested in an unsupervised approach towards mining of patterns in the data. These patterns can be simple ...
1
vote
0answers
14 views

Deep RL: Visualizing/Analyzing the gradient

I am testing different RL methods, and I know e.g that policy gradient method is supposed to have a high variance gradient which can cause trouble. I want to run a few different Deep RL algorithms, ...
0
votes
1answer
58 views

How to properly save and load an intermediate model in Keras?

I'm working with a model that involves 3 stages of 'nesting' of models in Keras. Conceptually the first is a transfer learning CNN model, for example MobileNetV2. (Model 1) This is then wrapped by a ...
0
votes
0answers
13 views

Perceptron with three input

A perceptron with two input is seen in two dimensions x1 is the x axis x2 is the y axis but if we had another entry x3 the graph would be in 3d and x3 would be the z axis? and if we had thousands of ...
0
votes
0answers
11 views

normalize output of a feed-forward ANN

My feed-forward neural-network is modeling (regression) a multi-channel loss function. The output of the network is a vector y (size 10) that describes the loss ratio of the input signal x for each ...
0
votes
2answers
20 views

Best way to build a wave classification system

I want to make a classifier for waves such as following: Above image is from: http://www.invisiblesbook.com/equal-temperament-tuning/ I believe, I will have to extract features from raw input using ...
2
votes
1answer
18 views

Classifying objects based of a varying number of the same type of feature vector for each object

For a congressional session, I have created a doc2vec model of speeches made. Using the vectors from this model, I have a dataset of each congressperson, their political affiliation, and a list of the ...
0
votes
0answers
8 views

Is it worth using residual blocks in a neural network with low number of layers?

I am new to the Deep Learning domain and I was recently reading about the resnet architecture. So I was wondering, can residual blocks improve the performance of even more "shallow" networks, or they ...
0
votes
1answer
16 views

Check Overfitting in CNN

I am kind of new to NLP and text classification with Convolutional Neural Nets, and I have trained my first models quite recently. I am a little bit concerned with overfitting. I am doing multilabel ...
0
votes
0answers
6 views

Why is it possible to train a semantic segmentation neural network like U-net/Tiramisu from scratch using small data-set like few hundreds

Why is it possible to train a semantic segmentation neural network like U-net/Tiramisu from scratch using small dataset like few hundreds. While for the classification task, it is not possible to ...