Skore: ImpurityDecreaseDisplay Aggregate Feature Added

Fourth PR, feature parity for CoefficientsDisplay in skore. :probabl. CoefficientsDisplay.frame() had the same gap as ImpurityDecreaseDisplay did before PR #2539: no way to aggregate coefficient values across splits. Added the same aggregate parameter pattern here, producing coefficient_mean and coefficient_std columns when set. The difference from the previous PR: this one also makes the relationship between frame() and _plot_matplotlib explicit. Plotting always passes aggregate=None so the visual output stays at per-split granularity regardless of what frame() defaults to. Keeps the two concerns cleanly separated. 4 commits, 97% code coverage, 1977 tests passing on merge. #opensource #python #skore #scikitlearn #machinelearning

Day 105 Backtracking is becoming more structured now. #Day105 🧩 131. Palindrome Partitioning How today went: • Iterated through the string, taking one substring at a time • Used DFS to explore partitions • Checked if each substring is a palindrome before going deeper • Built the result step by step and backtracked when needed What clicked: At each step: → choose a substring → validate (palindrome) → explore further → backtrack This pattern is repeating across problems. Backtracking is slowly becoming more intuitive. #LeetCode #DSA #Python #Backtracking #DFS #Recursion #LearningInPublic #Consistency

Day 109 Backtracking patterns are repeating again — and that’s a good sign. #Day109 🧩 78. Subsets How today went: • Used recursion to explore all elements • At each step, decide to include or skip the current element • Append current subset → explore → then pop to backtrack • Move to the next index and repeat What I’m noticing: Subsets is one of the cleanest backtracking patterns: → choose → explore → undo Another revision day, but clarity is improving. Consistency continues. #LeetCode #DSA #Python #Backtracking #Recursion #LearningInPublic #Consistency

Subsets: Classic Backtracking Template Generate all 2^n subsets via binary decision tree — include or exclude each element. Base case: index exceeds array length, save current subset copy. Backtracking: add element, recurse, remove element (backtrack), recurse again. Critical Detail: subset.copy() is essential — without it, all results reference same list, causing incorrect final output. Each subset snapshot must be independent. Time: O(2^n) | Space: O(n) recursion #Backtracking #Subsets #DecisionTree #DeepCopy #Recursion #Python #AlgorithmDesign #SoftwareEngineering

Subsets: Backtracking with Include/Exclude Decision Tree Generate all 2^n subsets via recursive binary choices — include current element or skip. Base case: processed all elements, save current subset. Backtracking pattern: modify state, recurse, undo modification. Backtracking Pattern: Modify shared state, explore branch, restore state before exploring alternate branch. This template applies to permutations, combinations, constraint satisfaction problems. Time: O(2^n) | Space: O(n) recursion depth #Backtracking #Subsets #DecisionTree #StateRestoration #Recursion #Python #AlgorithmDesign #SoftwareEngineering

Day 27/100 – DSA Practice Solved LeetCode 459: Repeated Substring Pattern today. Key Insight: If a string s is formed by repeating a substring, then it will always exist inside: (s + s)[1:-1] Why this works: Doubling the string generates all possible rotations Removing first & last characters avoids trivial matches If s is found in this modified string, it confirms a repeating pattern This elegant trick reduces the problem to a simple one-liner and highlights the power of pattern observation in strings. Approach Used: return s in (s + s)[1:-1] Constant progress > Perfection. On to the next one! #Day27 #100DaysOfCode #DSA #LeetCode #Python #CodingJourney #ProblemSolving

K Closest Points: Min-Heap for Efficient Selection Sorting all points costs O(n log n). Min-heap achieves same complexity but enables early termination — build heap with distances, extract k smallest. Squared distance avoids expensive sqrt while preserving ordering. Optimization Note: Max-heap of size k would be O(n log k) versus O(n log n) here. For small k, bounded heap beats full sorting. This solution works but isn't optimal for k << n. Time: O(n log n) | Space: O(n) #Heap #KClosest #DistanceCalculation #PriorityQueue #Python #AlgorithmDesign #SoftwareEngineering

Sixth PR in skrub, and the most interesting part of it wasn't the PR itself. PR: https://lnkd.in/eGyiGfcq While updating gallery examples to load datasets from file paths, I found that fetch_employee_salaries(split="test").employee_salaries_path was returning an empty CSV. Not something you'd catch by reading the code. Only came up because I was actually running the examples. Flagged it, maintainers confirmed. The path fix itself was routine, part of a broader effort (#1934) to standardize dataset loading across the example suite. One file I intentionally left untouched since it depended on fetch_figshare, which no longer exists. #opensource #python #skrub #scikitlearn #documentation

Day 111 Backtracking is starting to feel consistent now. #Day111 🧩 39. Combination Sum How today went: • Used backtracking with index i • Two choices: → stay at i (reuse the same element) → move to i + 1 (try next element) • Track the current total • If total == target → add result • If total > target → stop that path What I realized: This problem is about: → controlling index movement → managing the running total Same pattern, different control. Revision is making it clearer. #LeetCode #DSA #Python #Backtracking #Recursion #LearningInPublic #Consistency

📌 Problem: Reverse Vowels of a String 💡 Approach: Used the two-pointer technique to reverse only the vowels in the string. Initialize one pointer at the beginning and one at the end. Move both pointers inward until vowels are found, then swap them. Continue this process until both pointers meet. ⚙️ Key Insight: Use a set for fast vowel lookup (O(1)) Two-pointer approach avoids extra space for storing vowels separately ⏱️ Time Complexity: O(n) 📦 Space Complexity: O(n) (due to string → list conversion) 📚 What I learned: Efficient string manipulation using two pointers Optimizing lookups with hash sets #LeetCode #DSA #Algorithms #Coding #ProblemSolving #Python #TwoPointers #InterviewPreparation #CodingJourney