Skip to content

Instantly share code, notes, and snippets.

View anxingle's full-sized avatar

安兴乐-siler anxingle

View GitHub Profile
# fcn全连接网络+RNN
1. 切片为单列 第一层全连接中neuron为1024,batch_size:3000(太大)。 170epoch错误率 0.27/0.31
![enter description here][1]
第一层全连接neuron为512,batch_size:2500. 170epoch 错误率0.154/0.186;230epoch错误率0.154/0.186;280epoch错误率 0.063/0.078
![enter description here][2]
2. 切片为2列 第一层全连接中neuron为1024,batch_size:2500。280epoch错误率0.093/0.127.
![enter description here][3]
第一层全连接中neuron为512,batch_size:2500. 260epoch错误率0.089/0.12.
![enter description here][4]
3. 切片为3列 第一层全连接中neuron为1024,batch_size:2500.
# fcn全连接网络+RNN
1. 切片为单列 第一层全连接中neuron为1024,batch_size:3000(太大)。 170epoch错误率 0.27/0.31
![enter description here][1]
第一层全连接neuron为512,batch_size:2500. 170epoch 错误率0.154/0.186;230epoch错误率0.154/0.186;280epoch错误率 0.063/0.078
![enter description here][2]
2. 切片为2列 第一层全连接中neuron为1024,batch_size:2500。280epoch错误率0.093/0.127.
![enter description here][3]
第一层全连接中neuron为512,batch_size:2500. 260epoch错误率0.089/0.12.
![enter description here][4]
3. 切片为3列 第一层全连接中neuron为1024,batch_size:2500.
@anxingle
anxingle / smo.py
Created December 21, 2020 12:55
SMO
import numpy as np
import matplotlib.pyplot as plt
import time
from sklearn.datasets import make_blobs,make_circles,make_moons
from sklearn.preprocessing import StandardScaler
class SMOStruct:
""" 按照John Platt的论文构造SMO的数据结构"""
def __init__(self, X, y, C, kernel, alphas, b, errors, tol, eps, user_linear_optim, loop_threshold=2000):
@anxingle
anxingle / svm
Created December 22, 2020 01:29
smo data
# 生成测试数据,训练样本
X_train, y = make_blobs(n_samples = 1000, centers =2, n_features=2, random_state = 2)
# StandardScaler()以及fit_transfrom函数的作用需要解释一下
scaler = StandardScaler() #数据预处理,使得经过处理的数据符合正态分布,即均值为0,标准差为1
X_train_scaled = scaler.fit_transform(X_train, y)
y[y == 0] = -1
# 将训练数据绘制出来
fig, ax = plt.subplots()
ax.scatter(X_train_scaled[:, 0], X_train_scaled[:, 1],