Homec4science
Diffusion HDTorch (master) Browse

hdtorch/hdtorch/hyperdimGenerators.py
96ef2fc4baa2master

Last Change
Raw File

hyperdimGenerators.py
View Options

'''
Created on May 11, 2022
@author: wsimon
'''
import os, math
import torch
from . import _hdtorchcuda
def func_generateVectorsMemory_Random(numClasses, HD_dim):
''' function to generate matrix of HD vectors using random method
random - each vector is independantly randomly generated
'''
return torch.randint(0,2,(numClasses, HD_dim))
def func_generateVectorsMemory_Sandwich(numClasses, HD_dim):
''' function to generate matrix of HD vectors using sandwich method
sandwich - every two neighbouring vectors have half of the vector the same, but the rest of the vector is random
'''
vect_matrix = torch.zeros(numClasses, HD_dim)
for i in range(numClasses):
if i % 2 == 0:
vect_matrix[i, :] = torch.randint(0,2,HD_dim)
for i in range(numClasses - 1):
if i % 2 == 1:
vect_matrix[i, 0:int(HD_dim / 2)] = vect_matrix[i - 1, 0:int(HD_dim / 2)]
vect_matrix[i, int(HD_dim / 2):HD_dim] = vect_matrix[i + 1, int(HD_dim / 2):HD_dim]
vect_matrix[numClasses - 1, 0:int(HD_dim / 2)] = vect_matrix[numClasses - 2, 0:int(HD_dim / 2)]
vect_matrix[numClasses - 1, int(HD_dim / 2):HD_dim] = torch.randn(int(HD_dim / 2))
vect_matrix = vect_matrix>0
return vect_matrix
def func_generateVectorsMemory_ScaleRand(numClasses, HD_dim, scaleFact):
''' function to generate matrix of HD vectors using scale method with bits of randomization
scaleRand - every next vector is created by randomly flipping D/(numVec*scaleFact) elements - this way the further values vectors represent are, the less similar are vectors
'''
numValToFlip=floor(HD_dim/(scaleFact*numClasses))
vect_matrix = torch.zeros(numClasses, HD_dim)
#generate first one as random
vect_matrix[0, :] = torch.randint(0,2,HD_dim)
#iteratively the further they are flip more bits
for i in range(1,numClasses):
vect_matrix[i, :]=vect_matrix[i-1, :]
#choose random positions to flip
posToFlip=random.sample(range(1,HD_dim),numValToFlip)
vect_matrix[i,posToFlip] = 1-vect_matrix[i,posToFlip]
return vect_matrix
def func_generateVectorsMemory_ScaleNoRand(numClasses, HD_dim, scaleFact):
''' function to generate matrix of HD vectors using scale method with no randomization
scaleNoRand - same idea as scaleRand, just d=D/(numVec*scaleFact) bits are taken in order (e.g. from i-th to i+d bit) and not randomly
'''
numValToFlip = math.floor(HD_dim / (scaleFact * numClasses))
# initialize vectors
vect_matrix = torch.randint(0,2,(numClasses, HD_dim)).to(torch.int8)
# iteratively the further they are flip more bits
for i in range(1, numClasses):
vect_matrix[i] = vect_matrix[0]
vect_matrix[i, 0: i * numValToFlip] = 1 - vect_matrix[0, 0: i * numValToFlip]
return vect_matrix
def generateHDVector(vecType, numVec, HDDim, packed, device):
if vecType == 'sandwich':
hdVec = func_generateVectorsMemory_Sandwich(numVec, HDDim)
elif vecType == 'random':
hdVec = func_generateVectorsMemory_Random(numVec, HDDim)
elif "scaleNoRand" in vecType:
hdVec = func_generateVectorsMemory_ScaleNoRand(numVec, HDDim, int(vecType[11:]))
elif "scaleRand" in vecType:
hdVec = func_generateVectorsMemory_ScaleRand(numVec, HDDim, int(vecType[9:]))
else:
raise TypeError(f'Unrecognized vecType {vecType}')
if not packed:
return hdVec.to(device).to(torch.int8)
else:
hdVecPacked = _hdtorchcuda.pack(hdVec.to(device).to(torch.int8))
return hdVecPacked
c4science · Help