[ML]Self-supervised Learning
Compared to supervised learning, which requires both input data and labels, unsupervised learning, as its name suggests, only requires input data. Among vari...
Recent posts
[ML]Generative Adversarial Network
It’s common for neural networks to output values, vectors, or sequences for tasks such as classification or regression. Beyond this, networks can also be des...
[ML]Transformer: Attention is All You Need
In sequence-to-sequence (Seq2seq) problems, the model takes an input sequence and generates an output sequence, with the length of the output being determine...
[ML]Self-Attention
In machine learning models, the form of input can vary. It is not always a fixed-length vector, sometimes, it may be a set of vectors with varying lengths. F...
[ML][CNN]Convolutional Neural Networks
There are numerous network architectures tailored for specific tasks, and Convolutional Neural Networks (CNNs) represent a prominent class, specifically desi...