，不是真正的努力
最终是自己写个小程序解决的
最终是自己写个小程序解决的
&a href=&///?target=http%3A//blog.csdn.net/guotong1988/article/details/& class=& wrap external& target=&_blank& rel=&nofollow noreferrer&&spark pi 详解&i class=&icon-external&&&/i&&/a&
《数学之美》里面讲的不错
《数学之美》里面讲的不错
&a href=&///?target=http%3A///questions//relationship-between-rdd-partitions-and-nodes& class=& external& target=&_blank& rel=&nofollow noreferrer&&&span class=&invisible&&http://&/span&&span class=&visible&&/quest&/span&&span class=&invisible&&ions//relationship-between-rdd-partitions-and-nodes&/span&&span class=&ellipsis&&&/span&&i class=&icon-external&&&/i&&/a&
黑的应该是gfw，不是百度。李彦宏把吴恩达挖来，搞了研究院，他是一个追求技术懂技术的ceo。百度需要广告和外卖来生存，来养活研究院。搜索引擎不好做，失败例子很多，微软的都不行。谷歌也不是真的技术极高，也靠用户点击统计来提高搜索准确度。这么黑百度也是醉了，you can you up..
黑的应该是gfw，不是百度。李彦宏把吴恩达挖来，搞了研究院，他是一个追求技术懂技术的ceo。百度需要广告和外卖来生存，来养活研究院。搜索引擎不好做，失败例子很多，微软的都不行。谷歌也不是真的技术极高，也靠用户点击统计来提高搜索准确度。这么黑百度…
Yann LeCun在脸书主页上说&br&&br&&p&So much gee whiz hype in this article. Lots of people are training ConvNets through imitation learning (or reverse reinforcement learning) to drive cars.&/p&&p&Training a basic ConvNet to keep you in lane most of the time is fairly simple and straightforward. The problem is to make it work reliably.&/p&&p&The basic technique of training a neural net to keep you in lane was Dean Pomerleau at CMU in the late 1980s, a system called ALVINN. I used imitation learning to train a ConvNet for a self-driving robot called DAVE back in 2003. This work motivated the DARPA LAGR program.&/p&&br&&p&&b&What this guy is doing may be cool, but it isn't particularly innovative.&/b&&/p&
Yann LeCun在脸书主页上说So much gee whiz hype in this article. Lots of people are training ConvNets through imitation learning (or reverse reinforcement learning) to drive cars.Training a basic ConvNet to keep you in lane most of the time…
LeaderShip + 高级工程师 = 架构师
LeaderShip + 高级工程师 = 架构师
企业导师就是个P，三个字，靠自己
企业导师就是个P，三个字，靠自己
由CentOS转向Ubuntu后一切好了。。
由CentOS转向Ubuntu后一切好了。。
&a href=&///?target=http%3A//download.csdn.net/detail/tx122/6367229& class=& wrap external& target=&_blank& rel=&nofollow noreferrer&&prml课后习题答案完整版本&i class=&icon-external&&&/i&&/a&
哦，pppd是直接命令行里有的啊。。。
哦，pppd是直接命令行里有的啊。。。
他看到idea能马上反应出code该怎么写
他看到idea能马上反应出code该怎么写
我认为经常得小病，小病如果能调用全身的免疫器官的话，应该是不易得大病
我认为经常得小病，小病如果能调用全身的免疫器官的话，应该是不易得大病
&p&ListOrderedSet&/p&
ListOrderedSet
其他人说的理论，这些patch供参考&br&&br&SPARK-7076
Binary processing compact tuple representation&br&SPARK-7077
Binary processing hash table for aggregation&br&SPARK-7080
Binary processing based aggregate operator&br&SPARK-7081
Faster sort-based shuffle path using binary processing cache-aware sort&br&SPARK-7082
Binary processing external sort-merge join&br&SPARK-7083
Binary processing dimensional join&br&SPARK-7184
Investigate turning codegen on by default&br&SPARK-7190
UTF8String backed by binary data&br&SPARK-7251
Perform sequential scan when iterating over entries in BytesToBytesMap&br&SPARK-7288
Suppress compiler warnings due to use of sun.misc.Unsafe&br&SPARK-7293
Report memory used in aggregations and joins&br&SPARK-7311
Enable in-memory serialized map-side shuffle to work with SQL serializers&br&SPARK-7375
Avoid defensive copying in SQL exchange operator when sort-based shuffle buffers data in serialized form&br&SPARK-7440
Remove physical Distinct operator in favor of Aggregate&br&SPARK-7450
Use UNSAFE.getLong() to speed up BitSetMethods#anySet()&br&SPARK-7517
Rename unsafe module to managedmemory &br&SPARK-7691
Use type-specific row accessor functions in CatalystTypeConverters' toScala functions&br&SPARK-7698
Implement buffer pooling / re-use in ExecutorMemoryManager when using HeapAllocator&br&SPARK-7812
Speed up SQL code generation&br&SPARK-7813
Push code generation into expression definition
&br&SPARK-7814
Turn code generation on by default&br&SPARK-7815
Enable UTF8String to work against memory address directly&br&SPARK-7887
Remove EvaluatedType from SQL Expression&br&SPARK-7956
Use Janino to compile SQL expression&br&SPARK-8117
Push codegen into Expression&br&SPARK-8149
Break ExpressionEvaluationSuite down to multiple files&br&SPARK-8154
Remove Term/Code type aliases in code generation&br&SPARK-9700
Pick default page size more intelligently &br&SPARK-9548
BytesToBytesMap could have a destructive iterator&br&SPARK-9228
Combine unsafe and codegen into a single option&br&SPARK-9363
SortMergeJoin operator should support UnsafeRow &br&SPARK-9693
Reserve a page in all unsafe operators to avoid starving an operator&br&SPARK-9703
EnsureRequirements should not add unnecessary shuffles when only ordering requirements are unsatisfied&br&SPARK-8160
Tungsten style external aggregation &br&SPARK-9736
JoinedRow.anyNull should delegate to the underlying rows&br&SPARK-7165
Sort-merge Join for left/right outer joins
&br&SPARK-8159
Improve expression function coverage (Spark 1.5)&br&SPARK-8189
Use 100ns precision for TimestampType&br&SPARK-8190
ExpressionEvalHelper.checkEvaluation should also run the optimizer version&br&SPARK-8286
Rewrite UTF8String in Java and move it into unsafe package.&br&SPARK-8301
Improve UTF8String substring/startsWith/endsWith/contains performance&br&SPARK-8303
Move DateUtils into unsafe package&br&SPARK-8305
Improve codegen&br&SPARK-8307
Improve timestamp from parquet&br&SPARK-8317
Do not push sort into shuffle in Exchange operator&br&SPARK-8319
Update logic related to key ordering in shuffle dependencies&br&SPARK-8354
Fix off-by-factor-of-8 error when allocating scratch space in UnsafeFixedWidthAggregationMap&br&SPARK-8446
Add helper functions for testing physical SparkPlan operators &br&SPARK-8498
Fix NullPointerException in error-handling path in UnsafeShuffleWriter&br&SPARK-8579
Support arbitrary object in UnsafeRow &br&SPARK-8713
Support codegen for not thread-safe expressions&br&SPARK-8850
Turn unsafe mode on by default&br&SPARK-8866
Use 1 microsecond (us) precision for TimestampType&br&SPARK-8876
Remove InternalRow type alias in expressions package&br&SPARK-8879
Remove EmptyRow class&br&SPARK-9022
UnsafeProject&br&SPARK-9023
UnsafeExchange&br&SPARK-9024
Unsafe HashJoin&br&SPARK-9050
Remove out-of-date code in Exchange that was obsoleted by SPARK-8317&br&SPARK-9054
Rename RowOrdering to InterpretedOrdering and use newOrdering to build orderings&br&SPARK-9143
Add planner rule for automatically inserting Unsafe
Safe row format converters&br&SPARK-9247
Use BytesToBytesMap in unsafe broadcast join&br&SPARK-9258
Remove all semi join physical operator&br&SPARK-9266
Prevent &managed memory leak detected& exception from masking original exception&br&SPARK-9285
Remove InternalRow's inheritance from Row &br&SPARK-9329
Bring UnsafeRow up to feature parity with other InternalRow implementations&br&SPARK-9331
Indent generated code properly when dumping them in debug code or exception mode&br&SPARK-9334
Remove UnsafeRowConverter in favor of UnsafeProjection&br&SPARK-9336
Remove all extra JoinedRows&br&SPARK-9373
Support StructType in Tungsten style Projection&br&SPARK-9389
Support ArrayType in Tungsten&br&SPARK-9394
CodeFormatter should handle parentheses&br&SPARK-9411
Make page size configurable&br&SPARK-9412
Support records larger than a page size&br&SPARK-9413
Support MapType in Tungsten&br&SPARK-9418
Use sort-merge join as the default shuffle join Improvement RESOLVED
Reynold Xin &br&SPARK-9421
Fix null-handling bug in UnsafeRow.getDouble, getFloat(), and get(ordinal, dataType)&br&SPARK-9448
GenerateUnsafeProjection should not share expressions across instances&br&SPARK-9450
[INVALID] HashedRelation.get() could return an Iterator[Row] instead of Seq[Row]&br&SPARK-9457
Sorting improvements &br&SPARK-9464
Add property-based tests for UTF8String &br&SPARK-9738
remove FromUnsafe and add its codegen version to GenerateSafe&br&SPARK-9751
Audit operators to make sure they can support UnsafeRows &br&SPARK-9784
Exchange.isUnsafe should check whether codegen and unsafe are enabled&br&SPARK-9785
HashPartitioning compatibility should consider expression ordering&br&SPARK-9815
Rename PlatformDependent.UNSAFE -& Platform&br&　　更多关于Tungsten项目的进展，请关注SPARK-7075和SPARK-9697。&br&&a href=&///?target=https%3A//issues.apache.org/jira/browse/SPARK& class=& wrap external& target=&_blank& rel=&nofollow noreferrer&&Spark - ASF JIRA&i class=&icon-external&&&/i&&/a&
其他人说的理论，这些patch供参考SPARK-7076 Binary processing compact tuple representationSPARK-7077 Binary processing hash table for aggregationSPARK-7080 Binary processing based aggregate operatorSPARK-7081 Faster sort-based shuffle path …
多包容。。
多包容。。
1.打包&br&2.从多台机器读
1.打包2.从多台机器读
不是老板，但觉得老板应该好好反思自己的为人处事，而不是公司前景
不是老板，但觉得老板应该好好反思自己的为人处事，而不是公司前景
小公司看领导，&br&在小公司干了2.5年了，个人觉得领导很重要，&br&领导如果技术很高，人又很好，没有架子还行，&br&领导有技术，但是比较自傲、有派、不分享、只为自己产品考虑不管员工发展，就不好了，&br&但如果你觉得公司愿景不错，就大气点，想开点，打杂同时多学技术两手准备，干五年起不来就跳BAT&br&领导多少都有让你不爽的地方，如果又不是创始人之一，即使成功上市了也真心分成有限，所以怎么说都是干五年起不来就跳BAT，十年后拿个年薪百万，可是到5年的时候，自己投入那么多，舍得从头开始么
小公司看领导，在小公司干了2.5年了，个人觉得领导很重要，领导如果技术很高，人又很好，没有架子还行，领导有技术，但是比较自傲、有派、不分享、只为自己产品考虑不管员工发展，就不好了，但如果你觉得公司愿景不错，就大气点，想开点，打杂同时多学技术…}