Google App Engine で初めるServerSide JavaScripthagino 3000
This document discusses server-side JavaScript and mentions a Python hack-a-thon. It references CommonJS and server-side JavaScript, asks what something means, and mentions PHP and JavaScript can be used to create scalable web applications according to standards. A demo is also referenced.
This document is a profile page for @hagino3000, who works as a front-end engineer. It lists their interests which include iPhone Siri, Xbox Kinect, HTML5 device APIs, and various mobile platforms. It also recommends checking out the Emotiv brain-computer interface and provides links to its SDK and demos. The profile encourages hacking and experimenting with these technologies.
This document discusses Kinect and natural user interfaces (NUI). It provides information about Microsoft Kinect, including its use with Xbox 360 and potential support for Windows. It also mentions other companies working in this area like PrimeSense and their OpenNI/NITE software. Examples are given of hand and gesture recognition capabilities. Open source options like OpenCV are discussed for accessing Kinect data.
This document summarizes a presentation about hacking the Kinect motion sensing device. It discusses the drivers and libraries that allow accessing the Kinect's sensors from a computer, including the open-source libraries libfreenect and OpenNI. It also covers using the Kinect with openFrameworks and processing data from the Kinect in real-time using C++. Lastly, it discusses transmitting Kinect sensor data to a web browser over websockets using Node.js to enable controlling and visualizing the Kinect from a web page.
Kinect Hacks discusses the Kinect input device for Xbox 360, available for 13,000 yen in Japan. It describes open source drivers like libfreenect that enable Kinect use on non-Xbox platforms. Examples are given of Kinect being used for media art, game controllers, sex games, computer interfaces, and hand detection by MIT. APIs, documentation resources, and demo URLs are provided to help developers get started with Kinect hacks.
This document discusses Ext JS and Gears. It provides an overview of Ext JS including its use of JavaScript, UI components like grids and trees, and data binding using stores and proxies. It also discusses Ext Direct for remoting and Gears which introduced APIs in 2007 for features like databases, geolocation, and notifications to enhance HTML capabilities.
6. 例として正則化項を持つ最小二乗誤差関数を考える
N
1 2 λ T
∑{
J(w) =
2 n=1
T
}
w φ (x n ) − tn + w w
2
J(w)のwについての勾配を零とおく
↓
wについて偏微分すると零
(6.2
→
6.3の式展開)
7. N
d
J(w) = ∑{w φ (x n ) − tn }φ (x n ) + λ w = 0
T
dw n=1
wについて整理
1 N
w=−
∑
λ n=1
T
{w φ (x n ) − tn }φ (x n )
N
∑
= anφ (x n ) = Φ a
T
(6.3)
n=1
8. 6.2の式に
w=Φtaを代入する
N
1 2 λ T
J(a) = ∑{a Φφ (x n ) − tn } + a ΦΦ a
T T
2 n=1 2
ここで t
=
(t1,
…
tN)T とおくと
N
1 1 1 T
∑tn = 2 (t1t1 + t2t2 +... + tntn ) = 2 t
t
2 n=1
2
9. Σが外れて式6.5となる
1 T 1 T λ T
J(a) = a ΦΦ ΦΦ a − a ΦΦ t + t t + a ΦΦT a
T T T T
2 2 2
N*N対象行列のグラム行列 K = ΦΦT を定義
要素は
K nm = φ (x n )T φ (x m ) = k(x n , x m )
↑
6.1
のカーネル関数を利用する。
10. 6.5にグラム行列を代入
(6.5→6.7)
1 T T 1 T λ T
J(a) = a KKa − a Kt + t t + a Ka
2 2 2
二乗誤差関数をパラメータベクトルとカーネル
関数で表現できた → 双対表現
11. さらに式6.4からwを消去してaについて解いた
−1
a
=
(
K
+
λ
I
N
)
t
を線形回帰モデルに代入
T T T −1
y(x) = w φ (x) = a Φφ (x) = k(x) (K + λ I N ) t
予測値カーネル関数(と訓練データt)だけで表
現できた。 → 双対表現