Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Image recognition of handwritten digits in MNIS...
Search
Dimitris Spathis
January 17, 2016
Research
430
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Image recognition of handwritten digits in MNIST and flower classification in IRIS dataset
Code available here
https://github.com/sdimi/handwritten-digits-recognition
Dimitris Spathis
January 17, 2016
More Decks by Dimitris Spathis
See All by Dimitris Spathis
Song analysis "Jueves - La Oreja de Van Gogh "
sdimi
0
750
Wikipedia graph network analysis
sdimi
0
170
Character level LSTM Recurrent Neural Networks for language and music modeling
sdimi
0
320
Topic modeling and summarization of live TV shows via Twitter
sdimi
0
170
Diagnosing respiratory diseases with machine learning
sdimi
0
180
Detecting Irony on Greek Political Tweets: A Text Mining Approach
sdimi
0
130
Greece Unknown Facts
sdimi
1
190
Glocal News - Discover Top Local News, Globally.
sdimi
0
200
Responsive Web - Mobile First. An overview.
sdimi
1
190
Other Decks in Research
See All in Research
老舗ものづくり企業でリサーチが変革を起こすまで - 三菱重工DXの実践
skydats
0
180
SAKURAONE:An Open Ethernet-based AI HPC System And Its Observed Workload Dynamicsin a Single-Tenant LLM Development Environment
yuukit
1
320
NII S. Koyama's Lab Research Overview AY2026
skoyamalab
0
290
Ghost in the 7‑Zip: The Shadow of Residential Proxies Creeping into Your Life
nttcom
0
1.1k
姫路市 -都市OSの「再実装」-
hopin
0
1.7k
2026年3月1日(日)福島「除染土」の公共利用をかんがえる
atsukomasano2026
0
630
Can We Teach Logical Reasoning to LLMs? – An Approach Using Synthetic Corpora (AAAI 2026 bridge keynote)
morishtr
1
250
ブレグマン距離最小化に基づくリース表現量推定:バイアス除去学習の統一理論
masakat0
0
280
LLM Compute Infrastructure Overview
karakurist
2
1.4k
YOLO26_ Key Architectural Enhancements and Performance Benchmarking for Real-Time Object Detection
satai
3
800
FUSE-RSVLM: Feature Fusion Vision-Language Model for Remote Sensing
satai
3
860
社内データ分析AIエージェントを できるだけ使いやすくする工夫
fufufukakaka
1
1.1k
Featured
See All Featured
Primal Persuasion: How to Engage the Brain for Learning That Lasts
tmiket
0
360
Prompt Engineering for Job Search
mfonobong
0
340
Templates, Plugins, & Blocks: Oh My! Creating the theme that thinks of everything
marktimemedia
31
2.8k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
28
3.5k
Rails Girls Zürich Keynote
gr2m
96
14k
Helping Users Find Their Own Way: Creating Modern Search Experiences
danielanewman
31
3.2k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
38
2.9k
Odyssey Design
rkendrick25
PRO
2
690
What the history of the web can teach us about the future of AI
inesmontani
PRO
1
610
Introduction to Domain-Driven Design and Collaborative software design
baasie
1
830
Navigating the Design Leadership Dip - Product Design Week Design Leaders+ Conference 2024
apolaine
1
350
Digital Ethics as a Driver of Design Innovation
axbom
PRO
1
310
Transcript
Αναγνώριση χειρόγραφων χαρακτήρων στο MNIST dataset & φυτών στο IRIS
dataset με μεθόδους μηχανικής μάθησης Δημήτρης Σπαθής Εξαμηνιαία εργασία – Ιαν 2016 Υπολογιστική Νοημοσύνη – Στατιστική Μάθηση Καθ.: Αναστάσιος Τέφας
MNIST dataset 70.000 εικόνες 28 x 28 pixel 784 διαστάσεις
60.000 train 10.000 test {0,1,2,..9} multi-class
Εργαλεία Python python.org Scikit – learn scikit-learn.org Matplotlib matplotlib.org Numpy
numpy.org
Προεπεξεργασία δεδομένων Ανακάτεμα δειγμάτων X, y = shuffle(mnist.data, mnist.target) Κανονικοποίηση
pixels [0,1] X_train, y_train = np.float32(X[:60000])/ 255., np.float32(y[:60000])
Κρατάμε 90 components 90,3% της αρχικής πληροφορίας PCA – Μείωση
Διάστασης (784 → 90)
Εκπαίδευση SVM fitting classifier = svm.SVC(gamma=0.01, C=3, kernel='rbf') 5 Cross
validation cross_validation.cross_val_score(classifier, X_train, y_train, cv=5)
Αποτελέσματα εκπαίδευσης
Παραδείγματα ταξινόμησης
Μείωση Διάστασης Kernel PCA (784 → 300) kpca = KernelPCA(kernel="rbf",n_components=300
, gamma=1) LDA (300 → 9) lda = LDA() #should keep [classes – 1] components
Nearest Classifier K Nearest Neighbor clf = neighbors.KNeighborsClassifier(n_neighbors=5) Nearest Centroid
classifier = NearestCentroid(metric='euclidean', shrink_threshold=None)
Αποτελέσματα εκπαίδευσης
Embedding για Μείωση Διάστασης (784 → 2) Spectral Embedding manifold.SpectralEmbedding
(n_components=2, affinity='nearest_neighbors', gamma=None, random_state=None, eigen_solver=None, n_neighbors=5) Isomap Embedding manifold.Isomap(n_neighbors=5, n_components=2)
Spectral Clustering Kρατάμε 5000 δείγματα για οπτικοποίηση Spectral Clustering cluster.SpectralClustering(n_clusters=10,
eigen_solver='arpack', affinity="nearest_neighbors")
None
None
Αποτελέσματα clustering
IRIS dataset 150 λουλούδια 4 διαστάσεις sepal length sepal width
petal length petal width 3 κλάσεις Iris Setosa Iris Versicolour Iris Virginica
Αποτελέσματα SVM εκπαίδευσης
SVM fine-tuning C
SVM fine-tuning Degree
SVM fine-tuning Gamma
Μείωση Διάστασης
None
Αποτελέσματα clustering & embedding
None
None
Further work Kernel PCA – Memory Errors Incremental PCA Grid
Search Deep Architectures Distributed / Parallel MapReduce / Spark Κώδικας σύντομα στο github.com/sdimi