Conversation
src/alg_naive.py
Outdated
| @@ -0,0 +1,39 @@ | |||
| import tools as t | |||
|
|
|||
| def sp(n, k, L_t, d, unit_indices, results): | |||
There was a problem hiding this comment.
Naming: I'd give the function a longer and hence more descriptive name (e.g. shapley_value or compute_sv or something like that). Replace the single-letter names with more descriptive ones as well (e.g. n with num_units). All names should be small caps and underscores (e.g. L_t violates this, apart from violating the previous point as well).
src/alg_shapley.py
Outdated
| units are turned on at the start, and in the second case, 3 need to be on at the | ||
| start, so the computation of the DP table is different. | ||
| ''' | ||
| def sp(n_unit, k, L_t, Dp, d, uI, res): |
There was a problem hiding this comment.
Same as in the other file: function name, variable names, usage of capital letters...
tests/inspection/test_random.py
Outdated
|
|
||
| # Define the range for N we want to test and the number of repetition | ||
| # for each pair (N, K) | ||
| n_lower_bound = 2 |
There was a problem hiding this comment.
All constants should be in ALL_CAPS.
tests/inspection/test_random.py
Outdated
| N = n | ||
|
|
||
| # K from KNN classifier | ||
| K = k |
There was a problem hiding this comment.
These assignments seem uneccessary. Also K on the left is a variable I guess, so it should be named with small caps.
src/tools.py
Outdated
| return summ | ||
|
|
||
| 'Return a copy of the target vector' | ||
| def copy_vector(target): |
There was a problem hiding this comment.
This can be done with the copy built-in library.
src/alg_naive.py
Outdated
| # Compute the marginal contribution from the soft accuracy function | ||
| diff = prop_with_q-prop_without_q | ||
| vMv = diff/k | ||
| count = t.count_ones(vector) |
There was a problem hiding this comment.
Could be a simple single line, for example length([x for x in vector if x == 1]). In those cases it is excessive to put it in a function. Alternatively, you could use numpy to do these vector operations. It might be faster as numpy is highly optimized.
src/tools.py
Outdated
| tempScp.append(count(Di, D_c, 1, L_t)) | ||
| return [tempTb, tempSc, tempTbp, tempScp] | ||
|
|
||
| def minus(v): |
There was a problem hiding this comment.
These minus and sum functions could surely be replaced with numpy vector operations.
src/tools.py
Outdated
| return j | ||
|
|
||
| 'Compute n choose k' | ||
| def choose(n, k): |
src/tools.py
Outdated
| array : | ||
| array of tuple of the form [distance from D_t, label, id] | ||
| ''' | ||
| def sort(array): |
There was a problem hiding this comment.
This can be solved using the sorted built-in function with defining a custom key, see here.
2 participants
No description provided.