n = len(prices) if n < 2: return 0 # k is big enougth to cover all ramps. if k >= n / 2: return sum(i - j for i, j in zip(prices[1:], prices[:-1]) if i - j > 0) globalMax = [[0] * n for _ in xrange(k + 1)] for i in xrange(1, k + 1): # The max profit with i transations and selling stock on day j. localMax = [0] * n for j in xrange(1, n): profit = prices[j] - prices[j - 1] localMax[j] = max( # We have made max profit with (i - 1) transations in # (j - 1) days. # For the last transation, we buy stock on day (j - 1) # and sell it on day j. globalMax[i - 1][j - 1] + profit, # We have made max profit with (i - 1) transations in # (j - 1) days. # For the last transation, we buy stock on day j and # sell it on the same day, so we have 0 profit, apparently # we do not have to add it. globalMax[i - 1][j - 1], # + 0, # We have made profit in (j - 1) days. # We want to cancel the day (j - 1) sale and sell it on # day j. localMax[j - 1] + profit) globalMax[i][j] = max(globalMax[i][j - 1], localMax[j]) return globalMax[k][-1]