Questions tagged [embeddings]
The embeddings tag has no usage guidance.
185
questions
1
vote
0
answers
4
views
Why is T5 often used in text-to-data for text prompt encoders?
In the text-to-data(music, image, audio, etc.) generative AI field, one method of encoding text prompts is to use pre-trained language models. Such an approach was used in research on Moûsai [1] and ...
0
votes
0
answers
6
views
Combining Textual, Categorical and Numerical data for Semantic Search using SentenceTransformers model
I'm building a food semantic search model and I want to use a pre-trained SentenceTransformers model with cosine similarity. I'm using Epicurious dataset for the corpus which consists of textual (&...
0
votes
1
answer
22
views
Using text embeddings directly to compute similarity vs using them as features for a model that predicts similariy
Say you have a problem where you have a query and a set of result documents and you want to rank the result documents according to the query. Say also you have embeddings for the query and for the ...
2
votes
0
answers
25
views
CLIP Visual Transformer image encoder
I was doing some experiments with the CLIP's visual transformer encoder output (clip-ViT-B-32). So basically given the same scene or image, it should output almost ...
0
votes
1
answer
19
views
Are the terms "vector embedding" and "embedding vector" interchangeable?
Googling "vector embedding" gives me ~40k results, including from Pinecone, Weaviate, and OpenAI.
Googling ...
0
votes
0
answers
19
views
Are Vector embeddings position or direction vectors?
I've seen statements about embeddings being positions on a hypersphere. I've also seen statements about using dot product to determine similarity of these positions. This implies to me that we care ...
0
votes
0
answers
200
views
How to add knowledge to the LLM using LangChain (at a high level)?
At a super high level, I would like to create a fantasy language AI tutor. For this question, however, I would like to better understand how, generally speaking, you add your own custom data/knowledge ...
2
votes
2
answers
106
views
Why cant we use normalise position encodings instead of the cos and sine encodings used in the Transformer paper?
I'm working with Transformer models for sequence-to-sequence tasks and I'm trying to fully understand the use of positional encodings in these models.
In the original "Attention is All You Need&...
0
votes
2
answers
121
views
Recommended way to embed a text thousands of tokens long?
I've split the text up sections each 512 tokens long and created embeddings for each of them.
I want to combine them into 1 embedding for the full text. How do I do that? Is this even recommended?
...
0
votes
0
answers
35
views
Why are similarities all close to 1 in my two tower recommender?
I am trying to build a simple two tower recommender system on the MovieLens 100k dataset. The user tower is just a simple embedding layer. The item tower uses an embedding layer and concats that with ...
0
votes
1
answer
24
views
How do I interpret low dimentional embeddings of high dimentional embeddings?
I am trying to understand what I am supposed to learn about a problem when using dimensionality reduction methods. In particular, I am referring to methods like t-SNE and UMAP.
For the most part I am ...
0
votes
0
answers
12
views
High-Cardinality Categorical Feature with frequency score
I'm working with a really high-cardinality feature as one of the inputs to my model and I'm using hash-encoded feature embedding rather than one-hot encoding. However, this method is ignoring the ...
1
vote
0
answers
86
views
Otimization of similarity search for multiple embeddings by creating a weighted artificial embedding
I have embeddings of text created with a BERT model. A group of these embeddings should be used to find similar embeddings corresponding to this group. I know that you can use average or max (or ...
0
votes
0
answers
21
views
Max word embedding
I'm trying to understand the concept behind Max Word / Sentence embeddings also in contrast to mean word embedding. They used it for example in this article. So far I didn't find good resources which ...
0
votes
0
answers
26
views
Distance Metric for a dataset with embeddings and numerical columns
I'm trying to build an Approximate Nearest Neighbours model that can fetch similar records in a dataset, that are contextually similar. For example, for a record of job=Software Engineer and age=25, a ...
0
votes
0
answers
7
views
Align sub-sentences with sentences using embeddings
I would like ask for ideas
So I have sub-sentences, embedded in the context of their respective full-sentences. Then, I have other full-sentences and I would like to find
a) if they have similar sub-...
0
votes
0
answers
8
views
How to generate embedding for Image Dataset with multiple features?
My dataset has Jewellery images. The jewellery is of 4 types:
Type
Color
Shape
Material
Now I am aware of converting images to embeddings. I need to group similar images like two plain gold rings ...
0
votes
0
answers
13
views
How the embedding model (x-vectors) trained?
I read this paper: X-Vectors: Robust DNN Embeddings for Speaker Recognition which describes how PyAnnote embedding block works.
I'm not sure I understand how the X-Vector model was trained and tested:
...
0
votes
1
answer
141
views
SBERT Embeddings from Conversations
I have a dataset consisting of text-based conversations between two humans. One conversation has on average 20 turns and can look as follows:
...
0
votes
0
answers
9
views
In contrastive loss, why do we sample negatives from the document, not the query?
From what I read, for image search contrastive training, we give one query in text, and 1 positive image, and (N-1) sampled negative images.
However, for video search contrastive training, we give 1 ...
0
votes
0
answers
16
views
Does word2vec skip gram involves softmax in the output layer
I was going through various pytorch and from-scratch implementations of skip-gram. I found following:
This implementaiton does not seem to use softmax
...
1
vote
1
answer
32
views
Visualizing convolutional neural networks embedding
In this article, the author creates a graph (at the end of the post) from the embeddings of different words found by transformer model. I would like to do a similar thing for a convolutional neural ...
0
votes
0
answers
35
views
Inductive embedding technique for nodes without features
I am working on a project that involves graph embeddings. I have an homogeneous directed graph where the nodes have no features. The only available information are the edges between nodes. I need to ...
0
votes
1
answer
26
views
Is there a way to embed curves into smaller vectors?
I know about embeddings for words, but I would like to know if it is possible to do something similar for curves. What I mean by curves is a curve of a function. Say I have 1000 points corresponding ...
0
votes
0
answers
22
views
Select Topic Words from Clusters
I am following this solution for clustering: https://towardsdatascience.com/clustering-contextual-embeddings-for-topic-model-1fb15c45b1bd
For step four "Select Topic Words from Clusters", I ...
1
vote
1
answer
25
views
Which pre-trained model to select to generate embeddings from shop names written in English?
Good afternoon!
I have a dataset with thousands of shop names written in English. Several shop names might belong to one business entity, for instance, shops with names "KFC 001", "WWW....
0
votes
0
answers
21
views
Embedding: Can i use it in a time series problem?
I'm trying to do feature extraction in some discretized time series with a variable length, doing that i'm creating an RNN auto encoder. My main problem is to find a way to let the model train with ...
0
votes
1
answer
20
views
Are there methods to represent entire knowledge graphs using a single embedding vector?
In a knowledge graph, embedding vectors can be learned for nodes (node embedding) and edges (edge embeddings). Is there a method to learn one single embedding vector for the entire knowledge graph?
1
vote
0
answers
56
views
Machine learning with mixed variables dataset (numerical, categorical and embeddings)
I'm working on a machine learning project where I'm trying to predict the revenue of a movie.
My dataset contains mixed data types. There are numerical features (rating, number of votes, release year,....
1
vote
0
answers
77
views
Embedding performing unexpectedly worse than multi-hot encoding
I'm pretty new to ML and have set up a Keras model with a number of categorical features of users & items to predict positive reactions.
Each feature is a list of IDs from a database - i.e. ...
0
votes
0
answers
17
views
Embedding layers/entities in openAI's Hide and seek paper
I've recently come across a youtube video about openAI's hide and seek paper (https://openai.com/blog/emergent-tool-use/) and got really fascinated about the paper itself. But as I digging in the ...
0
votes
0
answers
11
views
hypothesis testing for difference between groups of embedding vectors
In a machine learning application, I have two generators G₁ and G₂ that generate descriptions of some input z in the form of embedding vectors X and Y:
Xᵢ = [x₁, x₂, ...] = G₁(zᵢ)
Yⱼ = [y₁, y₂, ...] =...
0
votes
0
answers
100
views
How to use a Swin Transformer with metric learning?
Using timm's implementation of Swin Transformer, how does one generate an embedding vector?
I would like to use timm's ...
0
votes
0
answers
9
views
How to find top k most co-related time series among a large set of time-series
Background: I am doing a project on epidemiological time series fore-castings, in which I am required to make predictions of a disease based on several symptoms.
About the dataset: The dataset of ...
0
votes
0
answers
24
views
Global threshold for clustering after training a Person-Reid neural network with triplet Loss?
Assume that I have "N" classes in my training data, I train a Person-Reid Neural Network using triplet Loss. In the inference stage I compute scores (using euclidean distance) as follows:
$$
...
0
votes
0
answers
38
views
Are there applications where you don’t need positional embeddings for transformers?
Are there applications where you don’t need positional embeddings for transformers?
Applications using positional embeddings with transformers: machine translation, image classification, etc.
0
votes
0
answers
52
views
Practical application of denoising autoencoders
I have been reading into autoencoders for the purpose of denoising data. In the examples i found (eg. [1, 2, 3], which are the first few google results) they have the following input/output:
Input ...
1
vote
1
answer
43
views
Understand the reason of embedding and the size inside it in Pytorch
I'm very new to pytorch - taking a course in udemy.
There is something I find hard to understand and would like to get explaination about, in a bit simpler words than what I can find in the ...
1
vote
1
answer
55
views
Biometrics identification with embeddings comparison and "unknown"/"other" class/label
This is a general or more conceptual questions about biometric classification models, based on deep learning neural networks.
The goal of the system is to take a set of features (e.g. voice recording, ...
1
vote
1
answer
67
views
Product embeddings
Hei, I have a list of purchase baskets from customers and would like to build embeddings for the products.
For example:
BASKET1 = ['PRODUCT234', 'PRODUCT214', 'PRODUCT768']
BASKET2 = ['PRODUCT2', '...
1
vote
1
answer
117
views
Is there a sensible notion of 'character embeddings'?
There are several popular word embeddings available (e.g., Fasttext and GloVe); In short, those embeddings are a tool to encode words along with a sensible notion of semantics attached to those words (...
1
vote
0
answers
187
views
How are the embedding and context matrices created and updated in word embedding?
I am struggling to understand how word embedding works, especially how the embedding matrix $W$ and context matrix $W'$ are created/updated. I understand that in the Input we may have a one-hot ...
2
votes
1
answer
2k
views
Transformer time series classification using time2vec positional embedding
I want to use a transformer model to do classification of fixed-length time series. I was following along this tutorial using keras which uses time2vec as a positional embedding. According to the ...
1
vote
1
answer
597
views
Do we perform text embedding before or after train-test splitting?
Do we perform text embedding before or after train-test splitting?
I know that for encoding variables, usually done after the split. However, I'm not sure if that's also the case for text processing?
2
votes
0
answers
14
views
Dicussion for X-vectors
I am posting this question to ask some questions regarding the X-vector embedding proposed by Synder et al.
The paper is X-VECTORS: ROBUST DNN EMBEDDINGS FOR SPEAKER RECOGNITION
With reference to the ...
1
vote
1
answer
382
views
What is the meaning of two embedding layers in a row?
I've noticed in one deep pre-trained textual neural network that there are two embedding layers in the beginning and I don't quite understand why there are two of them. As far as I understand (correct ...
1
vote
0
answers
69
views
Graph Neural Network | How node embeddings are learned from several graphs?
I am reading paper on MEGnet which is a GNN. The objective is that we have several molecules that share same elements such as molecules $C0_2$ and $COOH$ share $C$ and $O$. Now if we learn the node ...
1
vote
1
answer
1k
views
How can I get the output of a Keras LSTM layer?
I want to get the output (that is a vector) of a LSTM layer of a network built in Python using Keras and that is trained to classify sentences (i.e. sequences). How can I do it ?
My attempt has been ...
1
vote
1
answer
122
views
Latent space vs Embedding space | Are they same?
I am going through variational autoencoders and it is mentioned that:
continuity (two close points in the latent space should not give two
completely different contents once decoded) and completeness ...
2
votes
2
answers
172
views
Categorical Variable Embedding
I have a categorical variable in my labeled dataset. I trained one-hot encoded version of it in another neural network having embedding layer with a larger labeled dataset. I have obtained the weights ...