I'm working on implementation of LSTM Neural Network for sequence classification. I want to design a network with the following parameters:
Input : a sequence of n one-hot-vectors.
Network topology : two-layer LSTM network.
Output: a probability that a sequence given belong to a class (binary-classification). I want to take into account only last output from second LSTM layer.
I need to implement that in CNTK but I struggle because its documentation is not written really well. Can someone help me with that?
There is a sequence classification example that follows exactly what you're looking for.
The only difference is that it uses just a single LSTM layer. You can easily change this network to use multiple layers by changing:
LSTM_function = LSTMP_component_with_self_stabilization(
embedding_function.output, LSTM_dim, cell_dim)[0]
to:
num_layers = 2 # for example
encoder_output = embedding_function.output
for i in range(0, num_layers):
encoder_output = LSTMP_component_with_self_stabilization(encoder_output.output, LSTM_dim, cell_dim)
However, you'd be better served by using the new layers library. Then you can simply do this:
encoder_output = Stabilizer()(input_sequence)
for i in range(0, num_layers):
encoder_output = Recurrence(LSTM(hidden_dim)) (encoder_output.output)
Then, to get your final output that you'd put into a dense output layer, you can first do:
final_output = sequence.last(encoder_output)
and then
z = Dense(vocab_dim) (final_output)
here you can find a straightforward approach, just add the additional layer like:
Sequential([
Recurrence(LSTM(hidden_dim), go_backwards=False),
Recurrence(LSTM(hidden_dim), go_backwards=False),
Dense(label_dim, activation=sigmoid)
])
train it, test it and apply it...
CNTK published a hands-on tutorial for language understanding that has an end to end recipe:
This hands-on lab shows how to implement a recurrent network to process text, for the Air Travel Information Services (ATIS) task of slot tagging (tag individual words to their respective classes, where the classes are provided as labels in the training data set). We will start with a straight-forward embedding of the words followed by a recurrent LSTM. This will then be extended to include neighboring words and run bidirectionally. Lastly, we will turn this system into an intent classifier.
I'm not familiar with CNTK. But since the question has been left unanswered for so long, I can perhaps suggest some advice to help you with the implementation?
I'm not sure how experienced you are with these architectures; but before moving to CNTK (which seemingly has a less active community), I'd suggest looking at other popular repositories (like Theano, tensor-flow, etc.)
For instance, a similar task in theano is given here: kyunghyuncho tutorials. Just look for "def lstm_layer" for the definitions.
A torch example can be found in Karpathy's very popular tutorials
Hope this helps a bit..
Related
a superimposed display for train/val splits using StatisticsGen
Hi,
I'm currently using tfx pipeline inside kubeflow. I struggle to have StatisticsGen showing a single graph with train and validation splits curves superimposed, allowing better comparaison distributions. this is exactly how tfdv.visualize_statistics(lhs_statistics=train_stats, rhs_statistics=eval_stats, lhs_name='train', rhs_name='eval') behaves (see illustration 1), and I would like StatisticsGen to also provide a superimposed splits graph.
Thanks for any reference or help so that i can move forward.
Regards
You can use something like
# docs-infra: no-execute
# Compare evaluation data with training data
tfdv.visualize_statistics(lhs_statistics=eval_stats, rhs_statistics=train_stats,
lhs_name='EVAL_DATASET', rhs_name='TRAIN_DATASET')
From the tensorflow data validation tutorial
I am building a tool from scratch that takes a sample of text and turns it into a list of categories. I am not using any libraries for this at the moment but am interested if anyone has experience in this territory as the hardest part that I'm struggling with is building in sentiment to the search. It's easy to word match but sentiment is much more challenging.
The goal would be to take something like this paragraph;
"Whenever I am out walking with my son, I like to take portrait photographs of him to see how he changes over time. My favourite is a pic of him when we were on holiday in Spain and when his face was covered in chocolate from a cake we had baked"
and turn it into
categories = ['father', 'photography', 'travel', 'spain', 'cooking', 'chocolate']
If possible I'd like to end up adding a filter for negative sentiment so that if the text said;
"I hate cooking"
'cooking' wouldn't be included in the categories.
Any help is greatly appreciated. TIA 👍
You seem to have at least two tasks: 1. Sequence classification by topics; 2. Sentiment analysis. [Edit, I only noticed now that you are using Ruby/Rails, but the code below is in Python. But maybe this answer is still useful for some people and the steps can be applied in any language.]
1. For sequence classification by topics, you can either define categories simply with a list of words as you said. Depending on the use-case, this might be the easiest option. If that list of words were too time-intensive to create, you can use a pre-trained zero-shot classifier. I would recommend the zero-shot classifier from HuggingFace, see details with code here.
Applied to your use-case, this would look like this:
# pip install transformers # pip install in terminal
from transformers import pipeline
classifier = pipeline("zero-shot-classification")
sequence = ["Whenever I am out walking with my son, I like to take portrait photographs of him to see how he changes over time. My favourite is a pic of him when we were on holiday in Spain and when his face was covered in chocolate from a cake we had baked"]
candidate_labels = ['father', 'photography', 'travel', 'spain', 'cooking', 'chocolate']
classifier(sequence, candidate_labels, multi_class=True)
# output:
{'labels': ['photography', 'spain', 'chocolate', 'travel', 'father', 'cooking'],
'scores': [0.9802802205085754, 0.7929317951202393, 0.7469273805618286, 0.6030028462409973, 0.08006269484758377, 0.005216470453888178]}
The classifier returns scores depending on how certain it is that a each candidate_label is represented in your sequence. It doesn't catch everything, but it works quite well and is fast to put into practice.
2. For sentiment analysis you can use HuggingFace's sentiment classification pipeline. In your use-case, this would look like this:
classifier = pipeline("sentiment-analysis")
sequence = ["I hate cooking"]
classifier(sequence)
# Output
[{'label': 'NEGATIVE', 'score': 0.9984041452407837}]
Putting 1. and 2. together:
I would probably probably (a) first take your entire text and split it into sentences (see here how to do that); then (b) run the sentiment classifier on each sentence and discard those that have a high negative sentiment score (see step 2. above) and then (c) run your labeling/topic classification on the remaining sentences (see 1. above).
I'm using BigQuery for machine learning, more specifically the k-means method for an unlabeled dataset where I'm trying to find clusters.
I'd like to know if someone has discovered how the BQ ML initiates the centroids.
I already tried looking at the documentation but either there is nothing or I couldn't find it.
CREATE MODEL `project.dataset.model_name`
OPTIONS(
model_type = "kmeans",
num_clusters = 3,
distance_type = "euclidean",
early_stop = TRUE,
max_iterations = 20,
standardize_features = TRUE)
AS
(SELECT * FROM `project.dataset.sample_date_to_train`
)
The results differ a little each time I run.
Has someone experience with that subject?
For someone who is still looking for an answer, recently there has been an update on BigQuery ML about this topic. Two new paramaters have been added to the CREATE MODEL statement, i.e.:
KMEANS_INIT_METHOD
KMEANS_INIT_COL
Basically you can set your custom K observations (belonging to the data table) that will serve as initial centroids for your K-means algorithm. You can find the relative documentation at this link. Maybe it's not the most exciting solution to your problem, but it's still something you can work with if you need reproducibility.
If I had to guess, it probably uses a similar logic to TensorFlow (BQML might be using TF under the hood as it is). Random partitioning seems to be the TensorFlow default, so that would be my guess.
The reason you are seeing different results each time you train the model, is due to the random nature of the initial values assigned to the centroids. The K-means algorithm begins by randomly selecting a value(position) for the k number of centroids chosen. If you review this documentation it explains the exact process when using the K-means algorithm1.
I've been working a bit with neural networks and I'm interested on implementing a spiking neuron model.
I've read a fair amount of tutorials but most of them seem to be about generating pulses and I haven't found any application of it on a given input train.
Say for example I got input train:
Input[0] = [0,0,0,1,0,0,1,1]
It enters the Izhikevich neuron, does the input multiply a weight or only makes use of the parameters a, b, c and d?
Izhikevich equations are:
v[n+1] = 0.04*v[n]^2 + 5*v[n] + 140 - u[n] + I
u[n+1] = a*(b*v[n] - u[n])
where v[n] is input voltage and u[n] is a general recovery variable.
Are there any texts on implementations of Izhikevich or similar spiking neuron models on a practical problem? I'm trying to understand how information is encoded on this models but it looks different from what's done with standard second generation neurons. The only tutorial I've found where it deals with a spiking train and a set of weights is [1] but I haven't seen the same with Izhikevich.
[1] https://msdn.microsoft.com/en-us/magazine/mt422587.aspx
The plain Izhikevich model by itself, does not include weights.
The two equations you mentioned, model the membrane potential (v[]) over time of a point neuron. To use weights, you could connect two or more of such cells with synapses.
Each synapse could include some sort spike detection mechanism on the source cell (pre-synaptic), and a synaptic current mechanism in the target (post-synaptic) cell side. That synaptic current could then be multiplied by a weight term, and then become part of the I term (in the 1st equation above) for the target cell.
As a very simple example of a two cell network, at every time step, you could check if pre- cell v is above (say) 0 mV. If so, inject (say) 0.01 pA * weightPrePost into the post- cell. weightPrePost would range from 0 to 1, and could be modified in response to things like firing rate, or Hebbian-like spike synchrony like in STDP.
With multiple synaptic currents going into a cell, you could devise various schemes how to sum them. The simplest one would be just a simple sum, more complicated ones could include things like distance and dendrite diameters (e.g. simulated neural morphology).
This chapter is a nice introduction to other ways to model synapses: Modelling
Synaptic Transmission
I've been slowly going through the tensorflow tutorials, and I assume I will have to again. I don't have a background in ML but am slowly pushing my way up.
Anyway, after reading through the RNN tutorial and running the training code, I am confused.
How does one actually apply the trained model so that it can be used to make language predictions?
I know this is a terrible noobish and simple question, but I believe it will be of use to others, as I have seen it asked and not answered in a satisfactory way.
In general, when you train a model, you first do a forward pass, and then a backward pass. The forward pass makes a prediction based on your input data, and the backward pass adjust your model based on how correct your prediction was.
So when you want to apply your model, you just do a forward pass with your new data as input.
In your particular example, using this code, you can see how it's done by looking at how they run the test set, starting line 286.
# They instantiate the model with is_training=False
mtest = PTBModel(is_training=False, config=eval_config)
# Then they can do a forward pass
test_perplexity = run_epoch(session, mtest, test_data, tf.no_op())
print("Test Perplexity: %.3f" % test_perplexity)
And if you want the actual prediction and not the perplexity, it is the state in the run_epoch function :
cost, state, _ = session.run([m.cost, m.final_state, eval_op],
{m.input_data: x,
m.targets: y,
m.initial_state: state})