Questions tagged [meta-learning]
21 questions
8
votes
2 answers
What does the term episode mean in meta-learning?
Recall in meta-learning we have a meta-set which is a data-set of data-sets:
$$ D_{meta-set} = \{ D_n \}^N_{n=1} $$
where $D_n$ is data-set (or usually a task). Usually defined as a data sampled from a target function for regression or N classes for…
Charlie Parker
- 5,836
- 11
- 57
- 113
2
votes
1 answer
Bayes rule when data $D$ is split into two independent parts: $D_a, D_b$
In this machine learning paper Overcoming catastrophic forgetting in neural
networks, they present to you equation 1, the log of bayes rule:
$$ \log p(\theta|D) = \log p(D|\theta) + \log p(\theta) - \log p(D) \quad (1)$$
Where $\theta$ are the…
clam
- 198
- 2
- 6
2
votes
1 answer
Why only accuracy is used in meta and few-shot learning as evaluation parameters?
I was going through many state-of-the-art papers in Meta-learning and few-shot learning, and I found that almost all use "accuracy" as evaluation criterion. Unlike other domains like object detection, NLP and other machine learning domain which use…
Rambo_john
- 35
- 3
2
votes
1 answer
How do coordinate wise meta-learning optimizers update learner networks?
I was reading Optimization as a model for few shot learning and Learning to learn by gradient descent by gradient descent as I noticed both papers use something they call
coordinate wise optimzers
for what seems good reasons for efficiency.…
Charlie Parker
- 5,836
- 11
- 57
- 113
1
vote
1 answer
Why is my stacking/meta-learning not outperforming the best base model?
I have a dataset of around 10,000 rows, with 500 features, response variable is binary classification.
I split the features into 5 equally sized groups (based on subject matter expertise), and trained 3 different models (RandomForest, XGBoost, SVM)…
Vladimir Belik
- 980
- 1
- 10
1
vote
0 answers
What does it mean to "fine tune a MAML model for testing"?
I came accross this warning in the higher PyTorch library
# Crucially in our testing procedure here, we do *not* fine-tune
# the model during testing for simplicity.
# Most research papers using MAML for this task do an extra
# stage…
Charlie Parker
- 5,836
- 11
- 57
- 113
1
vote
1 answer
What does the batch norm layer for MAML (model-agnostic meta-learning) do for during inference?
I was mainly wondering if we should use the running statistics we used during meta-training or the batch statistics for the current task (during meta-evaluation).
Detailed thoughts (from git issue here…
Charlie Parker
- 5,836
- 11
- 57
- 113
1
vote
0 answers
Aggregating results from multiple unsupervised clustering studies
Is there a framework, similar to meta-analysis, to interpret or aggregate the results from multiple unsupervised clustering studies?
For example, take two studies on evaluating patients with heart failure to identify unique subgroups:…
cscswimmer227
- 11
- 4
1
vote
1 answer
Why there is theta in index of gradient symbol in gradient descent update formula for MAML?
In this MAML paper, they use following formula of gradient descent update (see page 3, algorithm 1):
$$
\varTheta '\ =\varTheta \ −\ \alpha \nabla _{\varTheta }\mathcal{L}_{\mathcal{T}_{i}}( f_{\varTheta })
$$
My question is, what is the reason of…
Jan Musil
- 291
- 2
- 9
1
vote
0 answers
How to train a meta-learning model?
In meta-learning we are suppose to train task ($\mathcal{T}_{i}$) with meta-train dataset ($\mathcal{D}_i^{\text{tr}}$). Say, we are training a 5-way ($N:5$) 2-shots $K:2$ meta-learning model with batch-size equals to 16, how do we build the data…
Eugene Liu
- 43
- 4
1
vote
1 answer
Multiple-domain adaptation vs multi-task learning
I am confused with the definitions of domain adaptation and multi-task learning.
I have K datasets, each with the same feature and label space and thus the same learning problem, but with different domain P(x,y).
For each dataset, I would like to…
Tim D
- 11
- 3
0
votes
0 answers
Feature Engineering for Meta-Learning?
I'm doing some stacked generalization/meta-learning.
In blogs and posts, I have only seen people take the level 1 predictions and just directly use them as features for a level 2 model (no feature engineering). However, I recognize that there is a…
Vladimir Belik
- 980
- 1
- 10
0
votes
0 answers
What is Fine Tune in "Matching Networks for One Shot Learning" paper
I was going through this paper and I understood the concept of the Meta-learning framework and using a few-shot technique.
But when I tried to interpret the results in "Table 1: Results on the Omniglot dataset", I couldn't understand the column…
Pathi_rao
- 1
- 1
0
votes
0 answers
How to prepare data for Few-shot learning with imbalance data?
I have highly imbalanced data of 2 classes. For example, 4000 samples where the number of positive class is 20 samples.
My idea is:
Train = 2000 samples (50%: 10 positive samples and 1990 negative samples).
Test = 1000 samples (25%: 5 positive…
0
votes
0 answers
Are Federated Learning and Model-Agnostic Meta-Learning the Same Thing?
I am currently reading this paper [1] and [2].
The author makes the claim that their Federated Learning scheme is similar to Model-Agnostic Meta-Learning?
They state:
Interestingly, FFL is similar to Model-Agnostic Meta-Learning (MAML)
in three…
Complicated
- 33
- 4