compression_quantizer.py

import numpy class Quantizer: def max_lloyd(self, nums, d): r = [] dl = len(d) l = len(nums) dic = {} for i in range(l): number = nums[i] for k in range(dl): if d[k] <= number < d[k + 1]: if k in dic: dic[k][0] += number dic[k][1] += 1 else: dic[k] = [number, 1] for i in range(dl - 1): r.append(dic[i][0] / (dic[i][1])) return r def d(self, r): l = len(r) d = [0] for i in range(1, l): d.append((r[i - 1] + r[i]) / 2) return d def iterate(self, a, di, it): maxmargin = di[-1] i = 1 r1 = [] while i <= it: print(f'Iteration:{i}') r1 = self.max_lloyd(a, di) print(f'r: {r1}') di = self.d(r1) di.append(maxmargin) print(f'd: {di}') i += 1 return r1, di def deq(self, nums, d, r): new = [] ld = len(d) l = len(nums) for i in range(l): ans = [] for j in range(ld): if d[j] <= nums[i] < d[j + 1]: ans.append(round(r[j], 5)) new.append(ans) return numpy.array(new) def mse(self, a, dq): l = len(a) s = 0 for i in range(l): s += pow((a[i] - dq[i]), 2) return s / l def uniqua(self, a, d): l = len(a) new = [] q = [] for i in range(l): for j in range(len(d)): if d[j] <= a[i] < d[j + 1]: new.append((d[j] + d[j + 1]) / 2) q.append(j) return new, q def semiqua(self, a, d): l = len(a) new = [] q = [] dic = {} for i in range(l): for j in range(len(d)): if d[j] <= a[i] <= d[j + 1]: if j in dic: dic[j] = (dic[j][0] + a[i], dic[j][1] + 1) else: dic[j] = (a[i], 1) for i in dic: dic[i] = int(dic[i][0] / dic[i][1]) for i in range(l): ans = [] ansq = [] for j in range(len(d)): if d[j] <= a[i] < d[j + 1]: ans.append(dic[j]) ansq.append(j) break new.append(ans) q.append(ansq) return numpy.array(new) def getdecisions(self, d, a): m = (numpy.min(a), numpy.max(a) + 0.001) sec = (m[1] - m[0]) / d return [m[0] + i * sec for i in range(d + 1)] def clustering_based(self, nums, k): l = len(nums) sec = l // k r = [] c = [] for i in range(k): r.append(nums[sec * i]) print(r, 'initiate!') for i in range(10): c = self.Cluster_assignment(nums, r) return r, c def Cluster_assignment(self, nums, r): l = len(nums) c = [] for i in range(l): number = nums[i] cpr = [] for j, k in enumerate(r): dis = abs(number - k) cpr.append((j, dis)) cpr.sort(key=lambda x: x[1]) c.append(cpr[0][0]) dic = {} for i in range(l): if c[i] in dic: dic[c[i]][0] += nums[i] dic[c[i]][1] += 1 else: dic[c[i]] = [nums[i], 1] for i in dic: r[i] = dic[i][0] / dic[i][1] return c def cde(self, r, c): dx = [] for i in c: dx.append(round(r[i], 5)) return dx if __name__ == '__main__': q = Quantizer() a = [0, 0.8, 0.8, 0.7, 0.8, 0.8, 0.7, 1, 1, 1, 1, 1.2, 1.2, 1.2, 1.23, 1.74, 1.75, 1.73, 1.83, 1.84, 2.2, 2.2, 2.4, 2.36, 2.38, 2.9, 2.8] d1 = [0, 3/4, 3/2, 9/4] print(len(a)) b = [0, 0.01, 2.8, 3.4, 1.99, 3.6, 5, 3.2, 4.5, 7.1, 7.9] d2 = [0, 2, 4, 6] d = [0, 0.7345833333333333, 1.450535714285714, 2.2332857142857145, 3] r = [0.4666666666666666, 1.0025, 1.8985714285714284, 2.5680000000000005] du = [0, 3/4, 3/2, 9/4, 3] #q.iterate(a, d1) #dq, uq = q.uniqua(a, du) nr, nc = q.clustering_based(a, 4) ans = q.cde(nr, nc) print(nr) print(nc) print(ans) print('MSE: ', q.mse(a, ans)) ''' print(dq) print(qu) dq, qu = q.semiqua(a, du) print(dq) print(qu) print('MSE:', q.mse(a, dq))'''