bfprt算法

标签：

如果数组arr的长度为N，排序之后自然可以得到最小的k个数，此时时间复杂度为排序的时间复杂度即O(N*logN)。本题要求读者实现时间复杂度O(N*logK)和O(N)的方法。

利用堆：

    public int[] getMinKNumsByHeap(int[] arr, int k) {
        if (k < 1 || k > arr.length) {
            return arr;
        }
        int[] kHeap = new int[k];
        for (int i = 0; i != k; i++) {
            heapInsert(kHeap, arr[i], i);
        }
        for (int i = k; i != arr.length; i++) {
            if (arr[i] < kHeap[0]) {
                kHeap[0] = arr[i];
                heapify(kHeap, 0, k);
            }
        }
        return kHeap;
    }

    public void heapInsert(int[] arr, int value, int index) {
        arr[index] = value;
        while (index != 0) {
            int parent = (index - 1) / 2;
            if (arr[parent] < arr[index]) {
                swap(arr, parent, index);
                index = parent;
            } else {
                break;
            }
        }
    }

    public void heapify(int[] arr, int index, int heapSize) {
        int left = index * 2 + 1;
        int right = index * 2 + 2;
        int largest = index;
        while (left < heapSize) {
            if (arr[left] > arr[index]) {
                largest = left;
            }
            if (right < heapSize && arr[right] > arr[largest]) {
                largest = right;
            }
            if (largest != index) {
                swap(arr, largest, index);
            } else {
                break;
            }
            index = largest;
            left = index * 2 + 1;
            right = index * 2 + 2;
        }
    }

    public void swap(int[] arr, int index1, int index2) {
        int tmp = arr[index1];
        arr[index1] = arr[index2];
        arr[index2] = tmp;
    }


利用bfprt算法：



    public int[] getMinKNumsByBFPRT(int[] arr, int k) {
        if (k < 1 || k > arr.length) {
            return arr;
        }
        int minKth = getMinKthByBFPRT(arr, k);
        int[] res = new int[k];
        int index = 0;
        for (int i = 0; i != arr.length; i++) {
            if (arr[i] < minKth) {
                res[index++] = arr[i];
            }
        }
        for (; index != res.length; index++) {
            res[index] = minKth;
        }
        return res;
    }

    public int getMinKthByBFPRT(int[] arr, int K) {
        int[] copyArr = copyArray(arr);
        return select(copyArr, 0, copyArr.length - 1, K - 1);
    }

    public int[] copyArray(int[] arr) {
        int[] res = new int[arr.length];
        for (int i = 0; i != res.length; i++) {
            res[i] = arr[i];
        }
        return res;
    }

    public int select(int[] arr, int begin, int end, int i) {
        if (begin == end) {
            return arr[begin];
        }
        int pivot = medianOfMedians(arr, begin, end);
        int[] pivotRange = partition(arr, begin, end, pivot);
        if (i >= pivotRange[0] && i <= pivotRange[1]) {
            return arr[i];
        } else if (i < pivotRange[0]) {
            return select(arr, begin, pivotRange[0] - 1, i);
        } else {
            return select(arr, pivotRange[1] + 1, end, i);
        }
    }

    public int medianOfMedians(int[] arr, int begin, int end) {
        int num = end - begin + 1;
        int offset = num % 5 == 0 ? 0 : 1;
        int[] mArr = new int[num / 5 + offset];
        for (int i = 0; i < mArr.length; i++) {
            int beginI = begin + i * 5;
            int endI = beginI + 4;
            mArr[i] = getMedian(arr, beginI, Math.min(end, endI));
        }
        return select(mArr, 0, mArr.length - 1, mArr.length / 2);
    }

    public int[] partition(int[] arr, int begin, int end, int pivotValue) {
        int small = begin - 1;
        int cur = begin;
        int big = end + 1;
        while (cur != big) {
            if (arr[cur] < pivotValue) {
                swap(arr, ++small, cur++);
            } else if (arr[cur] > pivotValue) {
                swap(arr, cur, --big);
            } else {
                cur++;
            }
        }
        int[] range = new int[2];
        range[0] = small + 1;
        range[1] = big - 1;
        return range;
    }

    public int getMedian(int[] arr, int begin, int end) {
        insertionSort(arr, begin, end);
        int sum = end + begin;
        int mid = (sum / 2) + (sum % 2);
        return arr[mid];
    }

    public void insertionSort(int[] arr, int begin, int end) {
        for (int i = begin + 1; i != end + 1; i++) {
            for (int j = i; j != begin; j--) {
                if (arr[j - 1] > arr[j]) {
                    swap(arr, j - 1, j);
                } else {
                    break;
                }
            }
        }
    }

    public void swap(int[] arr, int index1, int index2) {
        int tmp = arr[index1];
        arr[index1] = arr[index2];
        arr[index2] = tmp;
    }

bfprt算法

标签：

原文地址：http://blog.csdn.net/wangfengfan1/article/details/47682619