Introduction

The steps all voice assistants likely use:

1. First, the assistant must convert the speech to text.

2. The text is run through a natural language processing (NLP) step, which turns the words into numeric data.

3. Finally, there's a classification of the utterance - what people say to the intent - what they want the voice assistant to do.

   阅读全文 »

Unsupervised Learning --- Word Embedding

World Class : 对 word vector 进行聚类

Word Embedding：将每一个word project 到 High dimensional space 中，使得含义相近的 word 在空间的距离相近，且不同的dimension 代表不同的含义

阅读全文 »

公开课

Machine Learning 2021 Spring：

图像生成领域

Information Bottleneck Disentanglement for Identity Swapping（换脸）

Introduction

Rich semantic representations of text, and a simple linear classifier on top of the embeddings. What this architecture does is to capture aggregated meaning of words in a sentence, but it does not take into account the order of words, because aggregation operation on top of embeddings removed this information from the original text.

阅读全文 »

Self-Attention

常见的Network 架构 - Self-attention

Convolutional Neural Network

常用于图像领域

Introduction

近年来，自然语言处理（NLP）作为一个领域经历了快速增长，主要是因为语言模型的性能取决于其 "理解 "文本的整体能力，并且可以在大型文本库上以无监督的方式进行训练。因此，预训练的文本模型，如BERT，简化了许多NLP任务，并极大地提高了性能。

Adaptive Learning Rate

each parameters different learning rate

公开课

Machine Learning 2021 Spring：（p1-p20）

• machine learning还有deep learning的基本概念（机器学习类别：Regression Classification Structured Learning）