And now, let's practice tricky clusters in a row.

在，让我们连续练习一些复杂聚类。

It's about us learning to cluster. What does that mean?

这是关于我们学习聚类。 那是什么意思？

It is able to cluster cells according to their positions in such a multi-dimensional space.

它能够根据细胞在多维空间中位置对它们进行聚类。

So a non contrastive method, for example, like clustering is one non contrastive method.

因此，非对比方法之一，例如聚类，便是一种非对比方法。

They used an algorithm called DBSCAN, Density Based Spatial Clustering.

他们使用了一种名为DBSCAN（基于密度聚类算法）算法。

Type 3 neighborhood clustering involves looser packing.

第三类邻域聚类表为较为松散排列方式。

Okay, so it's clustering proteins that seem to act in concert.

好，所以这是将那些似乎协同作用蛋白质进行聚类。

Research on Lightweight Citrus Flowering Rate Statistical Model Combined with Anchor Frame Clustering Optimization[J].

轻量级柑橘开花率统计模型结合锚框聚类优化研究[J]。

If we look closely at this clump, we can see that it's sorted along two axes.

如果我们仔细观察这个聚类，我们可以看到它是沿着两个轴排序。

For example, this clustering phenomenon led geneticists to a gene associated with LDL cholesterol (the bad kind) and another related to inflammation.

例如，这种聚类象导致基因学家了一种与低密度脂蛋白醇(一种有害醇)有关基因，另一种与炎症有关。

So the cat and dog or two crops of this image, they belong to the same cluster.

因此，猫和狗或这幅图像两个裁剪部分，它们属于同一个聚类。

And then, for example, like clustering or self-distillation or other ways of doing it.

例如，可以采用聚类、自蒸馏或其他方法来实这一目标。

And basically crops that are coming from different images to be in different clusters.

基本上，来自不同图像裁剪部分应归入不同聚类。

Now, typically, if you were to do this clustering, you would perform clustering offline.

通常情况下，如果你要进行这种聚类，你会选择离线执行聚类操作。

These clusters are then connected to each other by an algorithm, and 113 million line segments , which reveal new conceptual relationships.

这些聚类随后通过算法相互连接，形成了1.13亿条线段， 揭示了新概念关系。

So as you're going through the data, you're actually computing these clusters online.

因此，当你遍历数据时，实际上是在线计算这些聚类。

What is suave and the main improvement proposed in the paper on supervised learning of visual features by contrasting cluster assignments?

在对比聚类分配视觉特征监督学习论文中，什么是suave以及提出主要改进是什么？

So all your k clusters have equal contribution to these n samples.

因此，所有k个聚类对这n个样本贡献是均等。

Our clustering is not like really hard clustering, it's soft clustering.

我们聚类并非严格硬聚类，而是软聚类。

And the idea basically is that you want the features from these two crops of an image to lie in the same cluster.

基本思路是，你希望同一张图像两个裁剪部分所提取特征落在同一个聚类中。