Questions tagged [training]

The tag has no usage guidance.

Filter by
Sorted by
Tagged with
1
vote
1answer
36 views

How does training a ConvNet with huge number of parameters on a smaller number of images work?

I have two questions: I am wondering why is that a very deep model such as VGG-16 which has approximately 138 million parameters (Source) can be used as a model to be trained on just 1.3 million ...
4
votes
1answer
354 views

How to implement clipping the reward in DQN in keras

How to implement clipping the reward in DQN in keras? especially how to implement clipping the reward? Is this pseudo code correct: ...
1
vote
0answers
64 views

Add training data to YOLO post-training

I've been playing around with YOLOv3 and obtaining some good results on the ~20 custom classes I trained. However, one or two classes look like they can use some additional training data (not a lot, ...
3
votes
1answer
273 views

What are the effects of clipping the reward in stability?

I am looking for stabilizing my results of DQN, I found clipping is one technique to do it but I did not understand it completely! 1- what are the effects of clipping the reward, clipping the ...
0
votes
1answer
34 views

What do I initialise each model in cross validation with in a multi-layer Perceptron?

So, as far as my understanding goes, cross-validation is used to determine the best model. I understand that once we determine the best model, we then train it on the entire dataset. I'm supposed to ...
1
vote
1answer
100 views

What is difference between final episodes of training and test in DQN?

What is difference between running in final episode of training mode and running in test mode in DQN? Is there any difference more than after training and tune the hyper-parameters, we test for one ...
1
vote
1answer
67 views

has number of output layer of DNN any effect in speed of find the optimal answer of DNN?

has number of output layer of DNN any effect in speed of find the optimal answer of DNN? For instance the more episodes is needed to train a DNN when the number of outputs is more? Is it correct?
2
votes
1answer
43 views

Minimum Neurons in Neural Network

I use a brute-force mechanism to determine optimal hidden layers/neurons by incrementing the layers/neurons by 1 up to some maximums and then picking the optimal counts from the best performing model. ...
2
votes
0answers
31 views

Maximum Layers in “any” Neural Network [duplicate]

I have about 6 months of experience in building and using Neural Networks with no prior/formal training. As I explore this field further, I see a lot of discussions about determining how many layers/...
1
vote
0answers
21 views

How to prepare future data for training

Assume I have a large data of ecommerce website sessions with user id key (a user can have multiple sessions with random time between them). The data is on S3 in json gzipped format. On some sessions ...
2
votes
3answers
52 views

Which is more important - stable training results or good test results?

Which is more important - stable training results or good test results? For instance, is obtaining an unstable training accuracy in different epochs, but good test accuracy better? Or is obtaining a ...
2
votes
0answers
120 views

How to resolve the instability of average reward per episode in training of DQN (Deep Q-Network)?

what is shown when average reward per episode in training is unstable? If there is big difference between average reward per episode and final reward by test section, what we can say? For ...
2
votes
1answer
159 views

Why does exploration in DQN not lead to instability?

Why does action exploration in DQN not lead to instability? I see in DQN algorithms, that it selects random actions even after some iterations. My question is how does this approach not lead to ...
2
votes
1answer
315 views

Constant validation loss & accuracy, training accuracy fluctuates

I am training a Squeeze-net model for binary classification of images. I have 79968 images for training (50:50 for and against) and 8892 images in the validation set. After 35000 iterations my ...
1
vote
1answer
155 views

when to use dot product and when to use the common product In neural networks?

I wanted to know when to use dot product and when to not , I also don't know when we must transpose an array and why should we , could someone help me to understand this ? If you could give me the ...
2
votes
1answer
65 views

Why would a validation set wear out slower than a test set?

On this page of Google's Machine Learning Crash Course, we find the following statement: "Test sets and validation sets "wear out" with repeated use. That is, the more you use the same data to make ...
2
votes
0answers
21 views

Force Neural Net to attempt to predict every class

I am training a (deep) neural net to classify approximately 60 different classes. The range of occurrences of each class in the dataset is wide, 3 orders of magnitude from the most represented to the ...
0
votes
1answer
51 views

Train and predict on different data structures

Is it possible to train a Machine Learning model with a set of the same data structure, but use another data structure as input in prediction/classification ? If so, how ? Let me elaborate and ...
1
vote
1answer
97 views

Train neural network for regression with negative samples

I have training samples which have have vector $\vec x$ as input and a vector $\vec y$ as output - both vectors have real (float) numbers $\in \mathbb R$ as entries. I want to train a neural network ...
2
votes
1answer
691 views

Why is predicted rainfall by LSTM coming negative for some data points?

I have used supervised learning with LSTM network using tanh activation function and 0.1 dropout for time series prediction.my loss='mean_squared_error', optimizer='adam'. The predicted time series is ...
2
votes
1answer
700 views

May the training set and validation set overlap?

May the training set and validation set overlap? Similarly, may the testing set and validation set overlap?
1
vote
0answers
25 views

Training Gaussian Restricted Boltzmann Machines with Noisy Rectified (nrelu or ssu) linear hidden units

I'm not sure how to implement this architecture. I'm following this thesis (pages 17-19) or this paper but I'm not sure how to train it. I want to use this to extract features from raw audio. I know ...
3
votes
1answer
616 views

Gumbel Softmax vs Vanilla Softmax for GAN training

When training a GAN for text generation, i have seen many people feeding the gumbel-softmax from the generator output and feed into the discriminator. This is to bypass the problem of having to sample ...
2
votes
0answers
331 views

Matlab: setting static iterations per epoch in a CNN

I'm building a convolutional neural network using Matlab's neural network toolbox. I have code designed to cross-train the network with different data sets, using the previous network's layers in ...
0
votes
1answer
19 views

Observing a set of features while training is same as seeing it one time. How?

If we use the same training example to train multiple times, it won't bring any change to desired theta values. Can someone provide intuition behind it or some theoretical explanation to back this? ...
3
votes
2answers
184 views

Image classification with different number of training image for each class

I'm trying to train neural network for image classification with 4 different classes: Cars (22k training examples) Building (8k training examples) Pedestrian (5k training examples) Trees (1k training ...
0
votes
2answers
37 views

How do I control for some patients providing multiple samples in my training data?

I have a dataset with 50 patients. The patients are tracked over many years with a set of a few thousand features measured at somewhat random intervals. I am trying to predict a certain outcome (this ...
1
vote
1answer
130 views

Training Encoder-Decoder using Decoder Outputs

I am trying to build an encoder-decoder model for a text style transfer problem. The problem is I don't have parallel data between the two styles so I need to train the model in an unsupervised ...
5
votes
1answer
1k views

What is the difference between tensorflow saved_model.pb and frozen_inference_graph.pb?

I've re-trained a model (following this tutorial) from the google's object detection zoo (ssd_inception_v2_coco) on a WIDER Faces Dataset and it seems to work if I use ...
3
votes
1answer
10k views

Number of features of the model must match the input. Model n_features is `N` and input n_features is `X`.

I am new to data science and trying get some results. I'm applying Decision Tree Classifier. When my train and test datasets' size are not equal I get an error `...
1
vote
0answers
47 views

How can I train a model to modify a vector by rewarding the model based on the modified vectors nearest neighbors?

I am experimenting with a document retrieval system in which I have documents represented as vectors. When queries come in, they are turned to vectors by the same method as used for the documents. The ...
1
vote
1answer
39 views

How handle the add of a new feature to the dataset?

Let's say that you trained a model (eg. Random Forest) on a dataset with ten features (or columns). Now you add one or more features to the dataset. You need the information brought from these new ...
2
votes
3answers
134 views

400 positive and 13000 negative: how to split dataset up (train, test, validation)

Working on a medical diagnostic convolutional neural networking problem, and it's not obvious (to me) how the dataset should be split up. Do I have enough data to split it in 3, or should I just have ...
2
votes
1answer
147 views

Should I eliminate all ID columns and similar columns from training data? [closed]

This is a basic question so bear my ignorance. I feel like they contribute collectively in no way to the target. This is for performance and accuracy. The target is polar (0,1).
22
votes
4answers
7k views

Is it always better to use the whole dataset to train the final model?

A common technique after training, validating and testing the Machine Learning model of preference is to use the complete dataset, including the testing subset, to train a final model to deploy it on, ...
7
votes
4answers
23k views

Train, test split of unbalanced dataset classification

I have a model that does binary classification. My dataset is highly unbalanced, so I thought that I should balance it by undersampling before I train the model. So balance the dataset and then ...
4
votes
2answers
68 views

Can I use the training rows multiple times while training with different labels attached to it?

If I have a data set where for every text message, same but two labels are given. It could be that only one label has been filled. To visualise this scenario in real life, one may classify the accent ...
2
votes
2answers
5k views

Determine useful features for machine learning model

I am working with a dataset with hundreds of features. I wish to create a simple machine learning model using 7-10 features from the original dataset. My question is this: What quantitative metrics ...
1
vote
0answers
25 views

keras evaulate method results vary with equal testset

I trained segnet on a dataset of remote sensing imagery. When I run model.evaluate a set of metrics is returned. When I compile the model again with the same ...
2
votes
2answers
84 views

Mathematical algorithms to generate labels for training data

What are the possible approaches when we need to train a model, but the training dataset is really small? (Assuming we have a lot of data, just not many data are labeled) I know a library from ...
3
votes
2answers
197 views

Why can't I choose my hyper-parameter in the training set?

Say I've divided the data into 3 parts: training, validation and test. I know for example, that in Neural Networks, the number of hidden layers is a hyper parameter. Why can't I train numerous NN ...
1
vote
2answers
113 views

problems during training a MLP type of network

I trained a neural network model, a MLP type of network, where the first several layers are 1-D convolution for processing sequence type of input. However, the training process looks like as follows, ...
2
votes
1answer
57 views

How to determine the number of forward and backward passes in deep learning (CNN)? [closed]

Is there a way to determine the number of forward and backward passes in the training of a neural network using python?
2
votes
1answer
720 views

WGAN is too slow what are some ways to tweak for speed ups?

I have implemented a vanilla GAN which gave good results very fast but it had a lot of mode collapse issue, because of this I learned about WGAN which suppose to fix this, in fact they claim they have ...
3
votes
1answer
1k views

Train loss vs validation loss

I have a few basic questions about tracking losses during training. If I am using mini-batch training, should I validate after each batch update or after I have seen the entire dataset? What should ...
2
votes
1answer
77 views

Isn't computing the “tractable error” in Restricted Boltzmann Machines (RBM) intractable?

Let $v \in \{0,1\}^M$ be the visible layer, $h \in \{0,1\}^N$ be the hidden layer, where $M$ and $N$ are natural numbers. Given the biases $b \in \Re^M$, $c \in \Re^N$ and weights $W \in \Re^{M \times ...
2
votes
1answer
56 views

Basic classification question

I am wondering how I can manage a test data after using PCA or normalization and another thing like that in the classification because our model works on the representation given by its input vectors. ...
0
votes
1answer
456 views

What if my validation set is worse than my training?

I am running a CNN, on the 1st epoch my training set accuracy is 15% and validation set is 12%, by the 51st epoch my training accuracy is 87% and validation set is 13%. What is happening? What does it ...
0
votes
1answer
1k views

Error in training a merged model in Keras

I attempted to merge a VGG-16 and ResNet-50 model in Keras to benefit from the combined feature representations toward a binary classification task. I was successful in building and saving the merged ...
3
votes
1answer
575 views

“other” class in Image classification

In the MNIST dataset, you have 10 defined classes, one for each digit. But you don't have a "not a digit" class. It seems that most image classification datasets are the same. But in a business ...