diff options
author | Yuchen Pei <me@ypei.me> | 2018-12-03 10:10:26 +0100 |
---|---|---|
committer | Yuchen Pei <me@ypei.me> | 2018-12-03 10:10:26 +0100 |
commit | 7d020bdb443414c054ac3624b0dd021df291e804 (patch) | |
tree | 7fabb629d774f7156961278583f8fae4faefc9d4 /posts | |
parent | bfb344527a1628a43fd10b71a4f034fd11c818d7 (diff) |
added section "evaluating shap"
Diffstat (limited to 'posts')
-rw-r--r-- | posts/2018-12-02-lime-shapley.md | 47 |
1 files changed, 42 insertions, 5 deletions
diff --git a/posts/2018-12-02-lime-shapley.md b/posts/2018-12-02-lime-shapley.md index 0e80c88..b6b38cb 100644 --- a/posts/2018-12-02-lime-shapley.md +++ b/posts/2018-12-02-lime-shapley.md @@ -77,7 +77,7 @@ feature contributions of supervised learning models locally. Let $f: X_1 \times X_2 \times ... \times X_n \to \mathbb R$ be a function. We can think of $f$ as a model, where $X_j$ is the space of $j$th feature. For example, in a language model, $X_j$ may correspond to -the count of the $j$th word in the vocabulary. +the count of the $j$th word in the vocabulary, i.e. the bag-of-words model. The output may be something like housing price, or log-probability of something. @@ -97,7 +97,7 @@ x_i, & \text{if }i \in S; \\ That is, $h_x(S)$ masks the features that are not in $S$, or in other words, we are perturbing the sample $x$. Specifically, $h_x(N) = x$. Alternatively, the $0$ in the \"otherwise\" case can be replaced by some -kind of default value (see the last section of this post). +kind of default value (see the section titled SHAP in this post). For a set $S \subset N$, let us denote $1_S \in \{0, 1\}^n$ to be an $n$-bit where the $k$th bit is $1$ if and only if $k \in S$. @@ -123,7 +123,7 @@ The LIME model has a more general framework, but the specific model considered in the paper is the one described above, with a Lasso for feature selection. -One difference between our account here and the one in the LIME paper +**Remark**. One difference between our account here and the one in the LIME paper is: the dimension of the data space may differ from $n$ (see Section 3.1 of that paper). But in the case of text data, they do use bag-of-words (our $X$) for an "intermediate" representation. So my understanding is, in their context, there is an @@ -134,6 +134,8 @@ with $f(r^{-1}(h_{r(x')}(S)))$. As an example, in the example of $X$ being the bag of words, $X'$ may be the embedding vector space, so that $r(x') = A^{-1} x'$, where $A$ is the word embedding matrix. +Therefore, without loss of generality, we assume the input space to be +$X$ which is of dimension $n$. Shapley values and LIME ----------------------- @@ -276,8 +278,8 @@ $$v(S) = f(x_S, \mathbb E_{\mu_{N \setminus S}} z_{N \setminus S}). \qquad (9)$$ It is worth noting that to make the modified LIME model considered in the previous section fall under the linear SHAP framework (9), we need -to make a further specialisation, that is, change the definition of -$h_x(S)$ to +to make two further specialisations, the first is rather cosmetic: we need to +change the definition of $h_x(S)$ to $$(h_x(S))_i = \begin{cases} @@ -285,6 +287,38 @@ x_i, & \text{if }i \in S; \\ \mathbb E_{\mu_i} z_i, & \text{otherwise.} \end{cases}$$ +But we also need to boldly assume the original $f$ to be linear, which in +my view, defeats the purpose of interpretability, because linear models are +interpretable by themselves. + +One may argue that perhaps we do not need linearity to define $v(S)$ as in (9). +If we do so, however, then (9) loses mathematical meaning. +A bigger question is: how effective is SHAP? + +Evaluating SHAP +--------------- + +The quest of the SHAP paper can be decoupled into two independent components. +The niceties of Shapley values and the choice of the coalitional game $v$. + +The SHAP paper argues that Shapley values $\phi_i(v)$ are a good measurement because they +are the only values satisfying the some nice properties including the Efficiency +property mentioned at the beginning of the post, invariance under permutation +and monotonicity, see the paragraph below Theorem 1 there, which refers to Theorem 2 of +Young (1985). + +Indeed, both efficiency (the "additive feature attribution methods" in the paper) +and monotonicity are meaningful when considering $\phi_i(v)$ as the +feature contribution of the $i$th feature. + +The question is thus reduced to the second component: what constitutes +a nice choice of $v$? + +The SHAP paper answers this question with 3 options with increasing simplification: +(7)(8)(9) in the previous section of this post (corresponding to (9)(11)(12) in the paper). +They are intuitive, but it will be interesting to see more concrete (or even mathematical) +justifications of such choices. + References ---------- @@ -307,3 +341,6 @@ References - Strumbelj, Erik, and Igor Kononenko. "An Efficient Explanation of Individual Classifications Using Game Theory." J. Mach. Learn. Res. 11 (March 2010): 1--18. +- 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>. |