论文修改建议 (LiY 20211108 细节问题)

1. Therefore, the added unlabeled data and pseudo-label both are key issues to improve the learning performance.
   $\to$ data 怎么是 key issue 呢？key issue 通常是一个任务、技术，而不是数据、标签。
   Therefore, the quality of both added unlabeled data and assigned pseudo-labels is critical to the learning performance.

2. In the regression domain, the co-training approach can still be effective in mining unlabeled data.
   $\to$ 我完全不懂你这里说的 still 是什么意思。是与 classification 相比？
   $\to$ 不是 domain, 是 task
   $\to$ 从这一句到本段末尾，有些啰里啰嗦的感觉，我也懒得去计较了
   Co-training is especially effective in dealing with the regression task.

3. Despite the advantages of co-training in SSR, there are still some limitations in the current methods.
   For co-training, most methods assume that unlabeled instances with high-confidence are credible for the model.
   $\to$ in 有点奇怪
   $\to$ 下文中你只指出了一个缺点, 这里不如直接说出来
   Despite the advantages of co-training in SSR, popular methods still have a major drawback.
   That is, they assume that unlabeled instances with high-confidence are credible for the model.

4. It is highly desirable to find a co-training algorithm that can reduce the impact of weak instances and labels.
   $\to$ 这是一个重要的结论，要把味道写出来
   $\to$ 下面的句子三选一
   Consequently, it is highly desirable to reduce the impact of weak instances and labels in the co-training process.
   Consequently, we need to reduce the impact of weak instances and labels in the co-training process.
   Consequently, it is highly desirable to reduce weak instances in the co-training process.

5. 你确定 F1 score 的 1 用下标？我看到的好像都没这样弄。还有 Top-k

6. in 3.1 这种没必要，在另外的小节才写成 in Section 3.1.

7. $g_j$ 可不可以换成 $g^{(j)}$? 这样和 $v^{(j)}$ 比较匹配的感觉。同时，$g_j(x_i)$ 就可以换成 $g_i^{(j)}$ 了。你自己考虑。

8. Eq. (5) can be regarded as a geometric projection problem as Figure 2 illustrates.
   $\to$ 前后相连，要写紧密些
   $\to$ 把式子里的句点改成逗号
   $\to$ 倒装一下
   which can be viewed as a geometric projection problem as illustrated in Figure 2.

9. We now present the practical details of SPOR through Algorithm 1.
   $\to$ 按照我的死规定，Figure, Algorithm, Table 必须是主语。由于下一段已经这样写了，这一句要直接删除。

10. Lines 5-6
    $\to$ 用两个横线，它会变成一个长的横线
    Lines 5–6