aboutsummaryrefslogtreecommitdiff
path: root/posts
diff options
context:
space:
mode:
authorYuchen Pei <me@ypei.me>2019-01-02 12:49:26 +0100
committerYuchen Pei <me@ypei.me>2019-01-02 12:49:26 +0100
commit9b5045d3363a009b44edb986a7482e7ec450a8ec (patch)
tree867ab88e16202b15ecf03d2a3dc90a174cbd3a97 /posts
parentd4731984b0162b362694629d543ec74239be9c73 (diff)
minor edit of lime-shapley post
- added note that the SHAP values are not original
Diffstat (limited to 'posts')
-rw-r--r--posts/2018-12-02-lime-shapley.md3
1 files changed, 3 insertions, 0 deletions
diff --git a/posts/2018-12-02-lime-shapley.md b/posts/2018-12-02-lime-shapley.md
index d21f594..a73d32e 100644
--- a/posts/2018-12-02-lime-shapley.md
+++ b/posts/2018-12-02-lime-shapley.md
@@ -265,6 +265,8 @@ value $\phi_i(v)$, where
$$v(S) = \mathbb E_{z \sim \mu} (f(z) | z_S = x_S). \qquad (7)$$
So it is a conditional expectation where $z_i$ is clamped for $i \in S$.
+In fact, the definition of feature contributions in this form predates
+Lundberg-Lee 2017. For example, it can be found in Strumbelj-Kononenko 2014.
One simplification is to assume the $n$ features are independent, thus
$\mu = \mu_1 \times \mu_2 \times ... \times \mu_n$. In this case, (7)
@@ -346,6 +348,7 @@ References
- Strumbelj, Erik, and Igor Kononenko. "An Efficient Explanation of
Individual Classifications Using Game Theory." J. Mach. Learn. Res.
11 (March 2010): 1--18.
+- Strumbelj, Erik, and Igor Kononenko. “Explaining Prediction Models and Individual Predictions with Feature Contributions.” Knowledge and Information Systems 41, no. 3 (December 2014): 647–65. <https://doi.org/10.1007/s10115-013-0679-x>.
- Young, H. P. “Monotonic Solutions of Cooperative Games.” International
Journal of Game Theory 14, no. 2 (June 1, 1985): 65–72.
<https://doi.org/10.1007/BF01769885>.