SVM Gamma ParameterYouTube↗  IA↗(long shot)
Education  34K views 195 likes 6 dislikes   17 Mar 2015
Udacity 7.7K subscribers N/A YouTube↗


Please select the problem
You must specify your email and describe the problem in detail
Describe your problem in detail if it's not obvious

Click to Play Video Click to Play Video
So what does the SVM γ parameter actually do?
إذن، ماذا تفعل معلمة جاما الخاصة بأجهزة متجهات الدعم؟
Well, on the sklearn documentation, it tells me
حسنًا، إنها تخبرني، في وثائق SKLearn،
that γ defines how far the influence of a single training example reaches.
بأن جاما تحدد مدى وصول تأثير أحد أمثلة التدريب.
If γ has a low value, then that means that every point has a far reach.
إذا كانت لجاما قيمة منخفضة، فذلك يعني أن كل نقطة لها وصول بعيد.
And conversely, high values means that each training example only
وعلى العكس ذلك، فإن القيم العالية تشير إلى أن لكل مثال من أمثلة التدريب وصول
has a close reach.
قريب فقط.
So what does this actually mean when we're fitting our support vector machine?
إذن، ماذا يعني ذلك عندما نقوم بملاءمة جهاز متجهات الدعم الخاص بنا؟
Well, suppose my data looks like this.
حسنًا، فلنفترض أن البيانات لدي تبدو كذلك.
All right, I think we all agree that the decision boundary is going to
حسنًا، أعتقد أننا جميعًا متفقون على أن حد القرار سيقوم
cut through the data something like this.
بتجزئة البيانات بطريقة ما مثل هذه.
The question is just exactly what this boundary is going to look like.
السؤال هو ماذا سيكون شكل الحد.
So if I have a high value of γ,
إذن، إذا كان لدي قيمة جاما مرتفعة،
the decision boundary, the exact details of the decision boundary,
فسيعتمد فقط حد القرار والتفاصيل الدقيقة الخاصة به
are just going to be dependent on the points that are the very closest to it.
على النقاط الأقرب إليه.
So these points that are very nearby,
وهي هذه النقاط القريبة جدًا،
which leaves it sort of effectively ignoring some of these points that fall
مما يعمل نوعًا ما على تجاهل بعض هذه النقاط التي تقع
a little bit further away from the from the decision boundary itself.
على مسافة بعيدة قليلاً من حد القرار نفسه.
On the other hand, if this is my decision boundary and
على الجانب الآخر، إذا كان هذا هو حد القرار الخاص بي
I have a low value of γ, then even the faraway points get taken into
وكانت لدي قيمة جاما منخفضة، فإذن ستتم مراعاة النقاط البعيدة أيضًا
consideration when I'm actually deciding where to draw the decision boundary.
عند قيامي باتخاذ القرار حول المكان الذي سأرسم فيه حد القرار.
Now, the reason why this matters, as you saw, is that, especially for
والآن، فالسبب وراء أهمية ذلك، كما شاهدت، وخاصةً
high values of γ, you can end up with a fairly wiggly decision boundary.
لقيم جاما المرتفعة، هو أنه يمكنك الحصول على حد قرار متلوٍ بعض الشيء.
So recall that when we had a high value of γ,
لذا، تذكر أنه عندما كانت لدينا قيمة جاما مرتفعة،
we got a decision boundary that might look something like this because, for
حصلنا على حد قرار قد يشبه شيئًا مثل هذا، ذلك لأن هذه النقطة
example, this particular point is going to be very nearby the decision boundary.
المحددة ـ على سبيل المثال ـ ستكون قريبة جدًا من حد القرار.
And so, it carries a lot of weight.
وبالتالي، فإنها ستحمل وزنًا كبيرًا.
It can pull the decision boundary all the way over so
يمكنها سحب حد القرار في كل المسافة بحيث
that it ends up on the correct side.
يصل إلى الجانب الصحيح.
On the other hand, if we have a low value of γ,
على الجانب الآخر، إذا كانت لدينا قيمة جاما منخفضة،
then these points that are near the decision boundary have a,
فسيكون لهذه النقاط القريبة من حد القرار
a relatively low weight when you compare them to the fact there's all these
وزنًا منخفضًا بشكل نسبي في حال مقارنتها بحقيقة وجود هذه
other faraway points that also influence the decision boundary.
النقاط البعيدة الأخرى جميعها التي تؤثر أيضًا على حد القرار.
So as we saw in this particular example, that ends up making the decision
وبالتالي، كما شاهدنا في هذا المثال المحدد، يعمل ذلك على جعل حد
boundary a little more linear, a little bit smoother, a little bit less jagged.
القرار مستقيمًا بشكل أكبر قليلاً ومستويًا قليلاً وأقل تعرجًا.
Similar videos:

Decision Tree Accuracy - Intro to Machine Learning
Udacity
25.2K views 2 likes 0 dislikes 23 Feb 2015100% match

Introduction Part II
Udacity
90.9K views 164 likes 12 dislikes 23 Feb 2015100% match

Bayesian Probabilities On Your Own - Intro to Machine Learning
Udacity
43.6K views 2 likes 0 dislikes 23 Feb 2015100% match

Class Labels After Second Split - Intro to Machine Learning
Udacity
26.4K views 6 likes 0 dislikes 23 Feb 2015100% match

Speed Scatterplot 3 - Intro to Machine Learning
Udacity
38.4K views 17 likes 2 dislikes 23 Feb 2015100% match

Confusion Matrices - Intro to Machine Learning
Udacity
28.3K views 27 likes 51 dislikes 23 Feb 2015100% match

Scatterplot to Decision Surface - Intro to Machine Learning
Udacity
34.5K views 14 likes 1 dislikes 23 Feb 2015100% match

Cancer Test - Intro to Machine Learning
Udacity
51.2K views 28 likes 6 dislikes 23 Feb 2015100% match

Scatterplots to Decision Surface - Intro to Machine Learning
Udacity
104.4K views 16 likes 0 dislikes 23 Feb 2015100% match

Entropy Calculation Part 2 - Intro to Machine Learning
Udacity
28.9K views 4 likes 7 dislikes 23 Feb 2015100% match

Why Is Naive Bayes Naive - Intro to Machine Learning
Udacity
64.5K views 58 likes 27 dislikes 23 Feb 2015100% match

Information Gain Calculation Part 4 - Intro to Machine Learning
Udacity
25.5K views 4 likes 1 dislikes 23 Feb 2015100% match

Linearly Separable Data - Intro to Machine Learning
Udacity
38K views 30 likes 14 dislikes 23 Feb 2015100% match

Unpacking NB Using Bayes Rule
Udacity
24K views 8 likes 1 dislikes 23 Feb 2015100% match

Posterior Probabilities - Intro to Machine Learning
Udacity
49.1K views 6 likes 0 dislikes 23 Feb 2015100% match

Min Samples Split - Intro to Machine Learning
Udacity
25.6K views 4 likes 0 dislikes 23 Feb 2015100% match

Separating Line - Intro to Machine Learning
Udacity
32.1K views 19 likes 1 dislikes 23 Feb 2015100% match

Bag of Words - Intro to Machine Learning
Udacity
50.5K views 88 likes 63 dislikes 23 Feb 2015100% match

Speed Scatterplot 3 - Intro to Machine Learning
Udacity
81.8K views 12 likes 3 dislikes 23 Feb 2015100% match

Features and Labels Musical Example
Udacity
69.5K views 98 likes 8 dislikes 23 Feb 2015100% match

Scatterplots to Predictions - Intro to Machine Learning
Udacity
34K views 8 likes 1 dislikes 23 Feb 2015100% match

Choosing Between Separating Lines - Intro to Machine Learning
Udacity
39.2K views 6 likes 1 dislikes 23 Feb 2015100% match

Speed Scatterplot 2 - Intro to Machine Learning
Udacity
107.5K views 12 likes 0 dislikes 23 Feb 2015100% match

Constructing a Decision Tree 2nd Split - Intro to Machine Learning
Udacity
26.5K views 6 likes 0 dislikes 23 Feb 2015100% match

Entropy Calculation Part 4 - Intro to Machine Learning
Udacity
28.8K views 2 likes 9 dislikes 23 Feb 2015100% match

SVM Response to Outliers - Intro to Machine Learning
Udacity
38.4K views 10 likes 2 dislikes 23 Feb 2015100% match

Bayes Rule for Classification - Intro to Machine Learning
Udacity
71.6K views 79 likes 36 dislikes 23 Feb 2015100% match

Classification By Eye Quiz - Intro to Machine Learning
Udacity
128.4K views 30 likes 0 dislikes 23 Feb 2015100% match

Coding A Decision Tree - Intro to Machine Learning
Udacity
33.8K views 9 likes 2 dislikes 23 Feb 2015100% match

Good Linear Decision Surface - Intro to Machine Learning
Udacity
37.1K views 31 likes 0 dislikes 23 Feb 2015100% match


Settings Privacy About Contact Donate