First Listings on Shenzhen Board Priced at High End

SHANGHAI — The first 10 companies due to list on ChiNext, the Nasdaq-style board in China, plan to sell shares at price-to-earnings ratios that are 50 percent above those of their main-board peers, just as worries over speculation spurred officials to announce strict new trading rules Thursday.

After gauging investor demand, the 10 start-up companies, including the software developer Beijing Ultrapower and the outdoor sportswear maker Toread, have decided on prices for their shares that average 55 times their 2008 earnings. That compares with an average price-earnings ratio of 36 for other initial public offerings this year on the mainland.

“Growth potential, rather than past performance, is what investors are looking at, so a high P/E ratio doesn’t necessarily mean they’re overpriced,” said Jiang Jianrong, an analyst at Shenyin & Wanguo Securities.

“But without doubt it will be quite a speculative market at the beginning, because it’s new and the companies are very small,” she added.

ChiNext may start operating as soon as next month in Shenzhen , the southern boomtown. To curb risks, the Shenzhen Stock Exchange, which will operate ChiNext, said Thursday that it would set an 80 percent limit on share-price movements during a stock’s first day of trade.

China is hoping that ChiNext can provide badly needed financing for private sector start-ups, which have difficulty obtaining bank loans but are crucial to creating jobs and sustaining growth. The long-planned market had until recently been referred to as the Growth Enterprise Market or the Growth Enterprise Board.

Beijing is also hoping that the market could become a cradle for China’s own future versions of Microsoft or Intel, helping to cut the economy’s reliance on manufacturing.

The 10 companies, which also include the drug producer Chongqing Lummy Pharmaceutical and Beijing Lanxum Technology, a provider of office information system services, will take subscriptions from investors starting Friday.

“Our rival Fuji Xerox is stronger than us both in branding and in financial strength,” said Lanxum’s chairman, Chi Yanming. “Listing on the second board would help us to narrow the gap.”

Lanxum, which is selling 5.3 million shares, said Thursday that it planned to raise 477 million yuan, or $70 million, 73 percent more than its previous fund-raising target, after pricing its I.P.O. at 18 yuan a share, or 51.49 times 2008 earnings.

Lepu Medical, a medical equipment maker, plans to raise 1.19 billion yuan, more than double its target, after pricing its I.P.O. at 29 yuan a share, or 53.54 times 2008 earnings.

Investor fervor is initially likely to push stocks on the start-up board to very high valuations, helping to create new Chinese billionaires.

“Some speculation is not always a bad thing,” Ms. Jiang said. “It provides easy money to private companies which had been at a disadvantage in financing compared with state-owned rivals.”

“More important,” she added, “the demonstration effect of start-up billionaires would prod more young Chinese to start their own businesses, and entrepreneurial spirit is what China desperately needs.”

But Chinese regulators apparently fear that the new board could become like a casino, with excessive speculation potentially hurting the interests of small investors and leading to a repeat of failures by similar boards in other countries created with start-up companies in mind.

Thirty-nine countries or regions have set up 75 such boards since the 1960s and nearly half have closed, a Shenyin & Wanguo Securities report said, including markets in Germany and Britain.

Under the price limits announced Thursday by the Shenzhen Stock Exchange, shares will be suspended from trade until the final three minutes of the session if they move more than 80 percent from the opening price on their first day of trade. This is in addition to debut-day circuit breakers that suspend trade for 30 minutes after price movements of 20 percent and 50 percent from the opening price.

深度学习是机器学习的一个子领域,它基于人工神经网络的研究,特别是利用多层次的神经网络来进行学习和模式识别。深度学习模型能够学习数据的高层次特征,这些特征对于图像和语音识别、自然语言处理、医学图像分析等应用至关重要。以下是深度学习的一些关键概念和组成部分: 1. **神经网络(Neural Networks)**:深度学习的基础是人工神经网络,它是由多个层组成的网络结构,包括输入层、隐藏层和输出层。每个层由多个神经元组成,神经元之间通过权重连接。 2. **前馈神经网络(Feedforward Neural Networks)**:这是最常见的神经网络类型,信息从输入层流向隐藏层,最终到达输出层。 3. **卷积神经网络(Convolutional Neural Networks, CNNs)**:这种网络特别适合处理具有网格结构的数据,如图像。它们使用卷积层来提取图像的特征。 4. **循环神经网络(Recurrent Neural Networks, RNNs)**:这种网络能够处理序列数据,如时间序列或自然语言,因为它们具有记忆功能,能够捕捉数据中的时间依赖性。 5. **长短期记忆网络(Long Short-Term Memory, LSTM)**:LSTM 是一种特殊的 RNN,它能够学习长期依赖关系,非常适合复杂的序列预测任务。 6. **生成对抗网络(Generative Adversarial Networks, GANs)**:由两个网络组成,一个生成器和一个判别器,它们相互竞争,生成器生成数据,判别器评估数据的真实性。 7. **深度学习框架**:如 TensorFlow、Keras、PyTorch 等,这些框架提供了构建、训练和部署深度学习模型的工具和库。 8. **激活函数(Activation Functions)**:如 ReLU、Sigmoid、Tanh 等,它们在神经网络中用于添加非线性,使得网络能够学习复杂的函数。 9. **损失函数(Loss Functions)**:用于评估模型的预测与真实值之间的差异,常见的损失函数包括均方误差(MSE)、交叉熵(Cross-Entropy)等。 10. **优化算法(Optimization Algorithms)**:如梯度下降(Gradient Descent)、随机梯度下降(SGD)、Adam 等,用于更新网络权重,以最小化损失函数。 11. **正则化(Regularization)**:技术如 Dropout、L1/L2 正则化等,用于防止模型过拟合。 12. **迁移学习(Transfer Learning)**:利用在一个任务上训练好的模型来提高另一个相关任务的性能。 深度学习在许多领域都取得了显著的成就,但它也面临着一些挑战,如对大量数据的依赖、模型的解释性差、计算资源消耗大等。研究人员正在不断探索新的方法来解决这些问题。
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值