Novas máquinas de autoaprendizagem baseadas em física poderiam substituir as atuais redes neurais artificiais e economizar energia
A inteligência artificial não só proporciona um desempenho impressionante, mas também cria uma procura significativa de energia. Quanto mais exigentes são as tarefas para as quais é treinado, mais energia consome.
![](data:image/jpeg;base64,/9j/4AAQSkZJRgABAQEAYABgAAD//gA7Q1JFQVRPUjogZ2QtanBlZyB2MS4wICh1c2luZyBJSkcgSlBFRyB2NjIpLCBxdWFsaXR5ID0gNjAK/9sAQwANCQoLCggNCwoLDg4NDxMgFRMSEhMnHB4XIC4pMTAuKS0sMzpKPjM2RjcsLUBXQUZMTlJTUjI+WmFaUGBKUVJP/9sAQwEODg4TERMmFRUmTzUtNU9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09P/8IAEQgCEgMgAwEiAAIRAQMRAf/EABsAAQACAwEBAAAAAAAAAAAAAAABAwIEBQYH/8QAGAEBAQEBAQAAAAAAAAAAAAAAAAECAwT/2gAMAwEAAhADEAAAAfTgAAAAAAA4XjvYePAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAH0b5z9GNgAAAAAAAAAAAAAAAHB8d7HxwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA+j/OPo5sAAAAAAAAAAAAAAAA4PjvY+OAAAAAAAAAAABtmo3dIAAAAAAAAAAAAAAAAAAAfR/nH0c2AAAAAAAAAAAAAAAAcHx3sfHAAAAAAAAAAAGx08be2q9Lf2TzQ45AAAAAAAAAAAAAAAAAAfR/nH0c2AAAAAAAAAAAAAAAAcHx3sfHAAAAAAAAAAC+jrVtYTPo3F2rBrc/0/mPPgIAAAAAAAAAAAAAAAAAfR/nH0c2AAAAAAAAAAAAAAAAcHx3sfHAAAAAAAAAAFnaov66uwjPpdO3Ka2eLu9DlPKJjlkAAAAAAAAAAAAAAAAB9H+cfRzYAAAAAAAAAAAAAAABwfHex8cAADYNd0RznRHOdEc50RznRHOdGTndPKvd2Lpjtq7CzGSimc9XO6jGNnznsNDjnzjp45nOdFHOdEc50RznRHOdEc50RznRHOdEc50RznRHOdEc50RznRoNUAAD6P84+jmwAAAAAAAAAAAAAAADg+O9j44AAbGvsG0TUJEJEAAJE30W6Z7OGx2zbTva2WhbdTtr7F2JX0uXsZc+j0PC5WpLNhIgBIhIhIgAAkhIhIhMCi+g1BAAD6P84+jmwAAAAAAAAAAAAAAADg+O9j44AAXU2myxJkxGTEZMYM2AsVi6zUyrq73G7HXlv6XT52sVJt3ypptusqmxNKL78b8vhu6PD0yxRmwVmwGbAZsILFaLFYzYDNgM2EmTEZU51lAUAB9H+cfRzYAAAAAAAAAAAAAAABwfHex8cAAL6NkuWrKloqi4UrhSuFK4U47Ip7fK71x0OdvU644V25XGEThYlBlMWSvM+q0cenzrYiboXooXiheKF6qFwpXClcKloqWipaKqdvXjTCgAPo/wA4+jmwAAAAAAAAAAAAAAADg+O9j44AAbOtsm2EAEkJEJEJEZRkbHpOZ1WdfGcXJOWJjllBCZWvLIkX0ZLzOV67iOvLTCgAAQkQmAAAABr7GuaYUAB9H+cfRzYAAAAAAAAAAAAAAABwfHex8cAANnW2TcEgAAAAC6vrnVxtomK5ymZxmC442Slc5wSiViQW0yczlet5Ot8dMKAAAAAAAA1tnXrSCgAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjgBMdlnjbLBd9ro2GuNhrl2GvJeoF6i5Nn0vN62VeEs5iMoicMpIRimZK4zkqK7IiJkRbWrncj1nM1rhrNXS5QL2uNhrjYa42GuNhrjYa42Naal1xQAD6P84+jmwAAAAAAAAAAAAAAADg+O9j44An0XnOoxGjFjVi5LStFS0VLRUtEdGr0OWxGVfKoRgziCccoIyDDKYM4NIEESATlhNanF9LXt5Grs6HWaq4UroKlwpXClcKVwpq26DUFAAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjgB1OX1GefbVa1uCEwJQJQFmHSl3OjXhx65xW57zVytinJMsq8YujXyW+aLLLq5p1LcIxxrKcYixVBdNE1fNE1fyuhlvn5nDucbtzwQ1kAAABr7GuaQoAB9H+cfRzYAAAAAAAAAAAAAAABwfHex8cDbTU6mjvM8+2q1vcglTAlCJmLFt7ups8u1lONfPrc1bJq6KszJWM8as5bYrWZ3VZXN+vsalmapndquC1VBdFQuUi7PWk2qJaxxau7ye3DXTGuZAlCpQidbY160xYAA+j/OPo5sAAAAAAAAAAAAAAAA4PjvY+OMu3w+wxr51Wpz7K7HTbYznUoGRks7uv0ufe+jWjl6srq6ZrZryrS9RlW/p233lqqKc9tm3Wts2M9ey89rT3NC5soswz1wyJrNhFlisWKxYrFtvPiXey1s9c9OjraHTzasW4dOMQhJQJ17qLNUWAAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjjLq8rqMae1q7Rz5iW70JZywg2LtGyb6ldO5x9uVeGzjvXGN6VWVWlNgW7uhu3jqY3YNa9k4zW4xy1x2Od0eerHPDPSIyhqEiIyGJUuVNl01XZOFzlrxsTWvdGv08+VNVPXyX11TrnnEEmqysoFgAD6P84+jmwAAAAAAAAAAAAAAADg+O9j442b7a2KdnX2E59ldjpssmdYMxXhfJrbtdeenZq1etx9WGvdfndVN1hXVfJVs123DX3tWzHDYxmoyZM3aO5RZhhfjN1RdEtSyFrwsyl1rdhZSzrKctuox1KtPrwt1dizfHUzvaxVOZMGYxo2aLNMWAAPo/zj6ObAAAAAAAAAAAAAAAAOD472Pjjb1BHU5fUTn21WtbYzQCQsrlc9iq/n16ddHQ49qosRVnbjVWWcmVd8axq5WRndeWcpFd2NleNyWmLYlqi7NaMrcUrrttXWsz5VlvN2p689bKKt8pxlrMJhAAGvsa9aYsAAfR/nH0c2AAAAAAAAAAAAAAAAcHx3sfHADqcvqM8+2q1rcEAAATs6uUvW2uV1eHS/JOLhOYxZqiZmyqLcZcWYwZSYRmiqc5jBnhWMWzLjrzFmnOOh1zZr4x1wFAgKCANbZ1zSFAAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjgB1OX1GefZXK7qlLe15L4pF00C9rybXS4V0evz4/T4yycCWQxtyyqzJY4lkYxJkxgziILIwxjPPDAsxx5tvQ4+lq9bt162Wl6gXxSLlQtVQXKRcpF1E1FAoAB9H+cfRzYAAAAAAAAAAAAAAABwfHex8cAOpy+ozz8sbVsXooXiheKF4ojYGvGyHovN7knoMq885yZYmOWMk4yIQJRiZSxRljjBXwbrZ5eWe9auewKI2Brtga7YGu2BrzeKF4oXiirc110xQAD6P84+jmwAAAAAAAAAAAAAAADg+O9j44AdTl9Rnn21WruCAAAAAMsq7jq73C7cxfGeNYRCTKIhJiaibIwSckK06dRqjLOi6QKAAAAAAAA1tnWrTCgAPo/wA4+jmwAAAAAAAAAAAAAAADg+O9j44AbWqTK/W2V3AgAAglAlAnLCTe6HH67HUwu1LMJqyvPOIEZ41JlbjWWaN2tN4acVzqCgAAAAAAAANbZ1jTCgAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjgABs615uqiWqpqxWLFYsVosVi1XlV3a4/obi/mTR6PDvRrZrdjgJtr10uzwrWOZlqef2WKU6WqhaqFqpFqoWqhaqFqoWqhaqFqoW6+dJrBQAH0f5x9HNgAAAAAAAAAAAAAAAHB8d7HxwAAtquLWZMGYwZwQkYshiyEW4X1td7Qx6YqUu/K7Y0bpdhREudtdBdp2crlqInLl0wZjBmMGYwZjBmMGYwZjBmMGYwZjBmMK76TXCgAPo/zj6ObAAAAAAAAAAAAAAAAOD472PjgABsa+wbaFSgSgAAATs0dHTc49mt2tqppaqHQa0ZZWxzcxlLlEwllAlAlAlAmAAAAAAAUX0GoIAAfR/nH0c2AAAAAAAAAAAAAAAAcHx3sfHAADY19g2RQAAACWRfbq6/Xdqpu2qhaqFu3zxZOUcs4wYyAAAAAAAAAAAAovoNUQAA+j/OPo5sAAAAAAAAAAAAAAAA4PjvY+OAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAH0f5x9HNgAAAAAAAAAAAAAAAHB8d7HxwAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA+j/OPo5sAAAAAAAAAAAAAAAA4PjvZeOISISISISISISISISISISISISISISISISISISISISISISISISISISISISISISISISI+j/OfoxsAAAAAAAAAAAAAAAA4XjvY+OAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAH0b5z9GNgAAAAAAAAAAAAAAAHC8cAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD6MGwAAAAAAAD//xAAsEAABBAEDAgYCAwEBAQAAAAAAAQIDBBQREhMgNAUhIiQyUBAxIzNgQBVB/9oACAEBAAEFAv8Ag8X7X/BV+3+g8X7X/BV+3+g8X7X/AAVft/oPF+1/wVft/oPF+1/wVft/oPF+1/6IoHyklZ7G/UV+3+g8X7X/AJ4Y+WSR+pGrmFmLjf8AT1+3+g8X7X/niTirMajG88mr2pYrfT1+3+g8X7X/AJomcki/yTO1ml0haR+h1yLjn+mr9v8AQeL9r/zVk2RN9EDv4o0heqeuJ07UsU/pq/b/AEHi/a/8rGq9701d5OmQc9zlbKQrwyW4eGf6Wv2/0Hi/a/8ALXbxRxelqJpE5u53K1DayUY5Y3Tw5FX6Wv2/0Hi/a/8AJFX0HKsr3+btfW5q6cbB7HRqj2yJE91Z92ojxU0+kr9v9B4v2vVEkartqm2qbaptqm2qbahtqG2obahpUNKh7RBLETDkWR2nC2JNjWN0STcLE9ox6tHRaox6tK+1W2NY11qKe0NtQ21TbVNtU21TbVNtU21TbUNtQ21DbUNtQ21DbUNtQ21DbUNtQ21DbVNtU21TbVNtUlSFE66/b/QeL9r1RMa9ceIx4jHiOCE4ITghOCE4ITghOGA4YDhgEjqoc8bEiTlX+yRnmsqq0a+RpsSQTexdWyHE9FRyzMkZW3cVcx4jHiMeIx4jHiMeIx4jHiOCI4ITghOCE4ITghOCE4ITghOCE4IjHiMeIx4jHiMeIx4iWJrE66/b/QeL9r1RROlXCkMKQwpDCkMKQwpDDeYbzDeYbzDcYbjDG14GiyN2t9EbE0HrsXVsg5rmLuR4sSnqQbM5CVkVmNasamG8wpDCkMKQw3mG8w3GG4w3GG4w3GG4w3mG8w3mG8w3mG8wpDCkMKQwpDCkMKQwpCWB0Sddft/oPF+16o43yLiTmJOYk5iTmJOYk5iTmJOYk5iTmJYMOwYVgSjMR1oohvrkb8XImqs8mv8AJ0YiqhyvOQZK1rrNNkyLSnQxJzEsGJYMSwYdgw7Bh2DDsGHYMOwYdgw7Bh2DDsGJOYk5iWDEnMScxJzEnMScxJzEnJIZI066/b/QeL9r1M36+4PcHuD3B7g9ye5NLBpYNLBpYNtg2TiQWFG0rCkESRNd6Wcb3Casc5qORHK03tU/iP4j+IjkbrPUnifpZNLBpYNLBtsG2wbbBtsG2wbbBtsGlg0sGlg0sGlg0sHuD3B7g9we4PcHuB/Jp11+3+g8X7XqjlfEuZYMywZdgzLBmWDLsGXYMuwZdgy7Bl2DLsGXYMmdTWypQZJpJoiL5qio40VjvQ44jiU4lONDc1pO100DpbjDLsGXYMuwZVgyrBlWDKsGVYMqwZdgy7Bl2DLsGXYMuwZdgzLBmWDMsGZYMywZlgzLBJPLKnXX7f6DxfteqGZYVzpDOkM6QzpTOlM6UzpTOlM6UzpTOlM6UzpTPmM6ypRWaQk013itTRNHIrNOjYp6WjfJba2WLnTGfMZ8xnzGfMZ8xnzGfMZ8xnzGdKZ0pnSmdKZ0pnSmdKZ0pnSmdKZ0pnSGdITWHTJ11+3+g8X7Xqhe2NcmEyYTJhMmEyYTJhMmEyYTJhMmEyYTJhMmIy2IR3ZXLHuSNU1NGqJq1XNEVUNxqhuNVNqIbdBjtS2+SJ2XGZUZlRmVGZUZlRGVEZURkwmTEZMJkwmTCZMJkwmTCZMJkwmTCZMJkwmTCTSskTrr9v8AQeL9r1QrEi7qRupG6kbqRupG+kb6RvpG+kb6RvpG+kb6QktNpUl5R3kmjnCtVD5ImrV9Kmw2G1puQ01NFaeThFUtJFE7kpHJSOSkb6RvpG+kb6RvpG+kb6RupG6kbqRupG6kbqRupG6kbqRupG6kTLAqddft/oPF+16oWxOXjpnHTOOmcdM46Zx0zjpmymbKZspmymbKZspDMBqx/BVF1URdBUPkKip+UapohorjVWmiKepBFRUswVWG2ibKRspGykbKRspHHTOOmcdM46Zx0zjpnHTOOmcdM46Zx0zjpnHTOOmTNganXX7f6DxfteqGNJFxGGIwxGGLGYsZixmLGYsZixmLGYsZixmNAVIayOcp5IblPkJ5CpoblNxuU1VRG+a6qa6m1TzQ3CaKlinXMaAxYzFjMRhiMMRhiMMRhiMMRhiMMRhiMMRhiMMRhiMMRhiMJoWxp11+3+g8X7XqhhfMuBOYE5gTmBOYE5gTGBMYExgTGBMYExgSkdBusUbYmKpoiHpFTQ+SIuh6VNqHpQ1RDVVFXz+R5oblNx6RCejC8Wg4wJjAmMCcwJzAnMCcwJzAnMCcwJzAnMCcwJzAnMCcwJzAnJq0kKddft/oPF+16KMWpPHxSsjfIuJYMSwYlgxLBiWDEsGJYMSwYlgxLBiWBtCw4q0mRKp+zabRF0F8l8nG1fxtU00E0N2iqhuU1Q1Q9J5HmT0YpR/h0yC07CGJYMSwYlgxLBiWDEsGJYMSwYlgxLBiWDFsGLYMWwYtgfDJGnXX7f6DxfteitMxxamZKMV6LyWDksG+wb7BvsG+wb7BvsG6wbrImW4hp2HK1qMYpqptU/R8hBU0NTcv401EQ8j4npNGnpNGmiGhqTVmyj6txguWhusG+yb7BvsG+wb7ByWDksHJYOSwclg5LByWDksD3SqnXX7f6Dxfteih/c/5xTPhXOsmdZM6yZ1kzrJnWTOsmdZM6yZtkjfflII3MaeRu/Gp+lU1NUPSamo08vxtNqm1TabTap6kNSRivSRl9g61cYZ1kzrJnWTOsmdZM6yZ1kzrJnWTOsmdZM6yZ1klsSzJ11+3+g8X7Xoof3P+cMvEua4zXGc4znmc8znmc8znmdINvTqsGS78eankhuN6n7EU1VF8lNFNFPJTXQRTQ+R5mv48zRTRx6jcpuQXXSVbrB12ZDOeZzzOeZrjNcZrjNcZrjNcZrjNcZriawsyddft/oPF+16KH9z/AJwvYxciuZFcyK5kVzIgMiAyIDJgMqIrzyyCeSK7Q9Tj0obmmrVFVWq5fLduF3IbhFVRdUPIb5N9TRf1yKcqnIpyKb1N6m9TkU5BHIakr50HWm65MBkQGRAZEBkQGRAZFcyK5kVzIrmRXJpInp11+3+g8X7Xoof3P+cPFrrSNaRrSNaRrSNaRupG+kRzV9zdVTU3taetxoekVNBr0UR2xzvSciocpyCSaCoI/wDjR+iquxd7FN0ZvYciHKcynMpzKcqG9gimpK5dHS1dd9I3UjWka0jWka0jWkeyPZHsibH066/b/QeL9r+G15HMdG5qUP7n/OFsbl4qpxVTiqnFVOKocdQ2UxMFCFW7VVERfM5Wob1cIxTa0amg/TTXka2Xackan8anI1gjmOPi5VRsSO3iPRDa1TYaNP4jdEckZyMOVhyMN8R/Eom43IP82uwtdtI46Zx1DiqnFVOKqcVU4qpxVSZkLU66/b/QeL9qJ+9HLPoradD+5/zij5FxDEQxEMRDFaYsYleshE2DX/66RGnm9drWHMep5xaCMQavm7+KTWJ5x6mx2v8AGwbH/KjV3fGJr+QXR4rHIaKcbziecZtjP4T+E1hP4TSI2IfytNzNXJuR0dRVx66mNGYrDFaYiGIhiGISw8addft/oPF+1/DWuiJkTSh/c/5xRPlXCnMKcwpzCmMGUwhteu0T0tfJqMjVwsiNRrHPX+OMWV7hInqcSDdzCRqPYsT0F1Qa6Rxxs1iVzSNHo13qakiSK71GkzDdObZlOF5xIbIzbEaQm2I42nC4VjmiOVDmURY1HNVyOgrqYbTCeYUphTGFMYU5hTksEkSddft/oPF+1E/c0KOfJEjG0P7n/NjXvOCc4JzHnMWwJSnEpog3iiFVXK2NGo+RXDI/J8uqMiVU5GsPXIvDoMa5omit2OFWRE3oiJFoKiboXK6R6at3sc5dDZM0/nOOZTgecKHGw42HE04FFhkQ80ElehyMccSOHNVoiqhybhYK7haSi0pzFsGNOY85wTj45GJ11+3+g8X7UT9zQpI+SHY2h/c/5sc9pzTnLOck5vnP51EisqNqzDdkQqq5UYjEe9XqjEYjnukVI2sR0qqJE5RI/NikjWq7SZosioqMVx5jH/zL8V4kcvkcb0NJzilU4HnCcJwqcDxWuQRzkOZTSJ46JzfwkrkHNhkFqyCwWUNk6H8xvnOSc5JzlnHvkcnXX7f6DxftTRvDrVJODbQ/uf8AOKV8S5s5mzmbOZk5lzmTMoj5nEcFlw3bEi+pUa2I83u9MQjVkd6Iza6ReJo3VCRE0SM1kabVkERUajtj1/SsiRfinEccpwvOBTgU4HCxOQ/QkjkNzHHEiit0GqrRVY8dXkUdHaaK+VBLEqGVOZk5mTmZOZs5mzktiSVOuv2/0Hi/ajI+SpiqSQcbaH9z/nFLxLlmWZZlmYpmvM+YW3KpFkyiaRpt3LokaNj3K5U0bGmiruOJEGt2/jZGojHoKx7xGIfATaLG002JxIpxPOFxwOOFxsVBFVDeptjcLFoaCOHx7h6WmGRIJcmQz5jNkMxTMUzDLMsyyWblTrr9v9B4v2or0xvxQ/uf84XRtXmrHNWOasc1Y5qxz1jJgEutaLfVSGSSZfJqIzcrv01gur1+BxiNTUciHGhxuNrD4p+xWamitONFOJxxuNjjzQ3ONUU40UVmgmrT0uJUlYLbVBL6oLf1MqFTIrnPXOasc1Y5qxzVjmrHNWJnwuTrr9v9B4v2vRQ/uf8AOHi19keyPZHsj2R7I9ia0Rs9Not9hBLzqvkjWDtXLptRrTarl2tNFNDa02IaN/G1wrdxorTYinG42uPUauNGqLGptEVRWeT7KRrnMQW9C4WWkprRPYnsT2R7I9keyPZHsj2RNj6ddft/oPF+16KH9z/nCxj14K5wVzgrHBWOCscFY4Kxw1TipjP/AD2GZBpC9swoiaIiaqvqVfM8uj0/naaK482m3U0chq41NEUVqoeaH7JXJEmZALcruR3/AJ7hYqRw1ThqnBWOCscFY4KxwVjgrHBXOCucFcmjiYnXX7f6Dxfteih/c/5wxcy4SmEphKYSmEphGEYRhGG0jrU2HPCiM0cn7VfI/SfpPJBPx5mpqp6lNEP2fo0NVNTT8IppqLLG1ciIkZSkHVIDDQwjCMIwjCUwlMJTCUwlMJTCUwlJq6xJ11+3+g8X7Xoof3P+cUL5lwbJg2TBsmDZMGyYNkwbJg2TAsmBZE8PnUh8PjYp+hD9qfo8l/PqPUaKeR+zU/RrqafjXUVPxJHHKj/DWqL4dOhgWTAsmBZMCyYNkwbJg2TBsmDZMGyYNkwbJg2TBskteWFOuv2/0Hi/a9FD+5/zYj1XjsHHYOKwcVg4rBxWDisHFYOKwcVg4JxtSy4r02xL+1XyP038ap+FNDQ0PSbj5Iiioa6Hk78ak1KOQfRsNFrzocM5xWDisHFYOKwcdk47Jx2TjsnHYOOwcdg47Bx2DjsD2yonXX7f6Dxfteih/c/5xyPjXKsGVYMqwZVgybBk2DJsGTYMmwZNgybBz2FI23JFgjWFrfMTzd+/xuE/HpPSaNPSatPiqiLoKn4d62y1rTRzrLTnmMmcyrBk2DKsGVYMqwZVgyrBlWDKsGVYMqwZVgyrBlWDKsD5pJE66/b/AEHi/a9FD+5/zhmfCufOZ85nzmfOZ85n2DPsGfYM+wZ9gz7Bn2DOsKQOuTL+k/Tfxq4/f4VVPSek2ioqHyai6CoIuhK170lluRLmWDOsmfYM+wZ9gz5zPnM+cz5zPnM+cz5zPnM+cz5zPnM+cz5zPnJrMkyddft/oPF+16KH9z/nDIka5bDLYZbDLYZbDLYZbTLaZbTLaZjTMaJeVD/0ZVKzpniebkFPUpqfo82mrVNv4ap+lcIuhYWZhnyH/oymbqZjTMaZjTLaZbTLaZbTLaZbTLYZbDLYZbDLYZbDLYZbDLYTTNkTrr9v9B4v2vRQ/uf84XRNXkpnJTOWoctQ5ahy1DlqHLUOWoctQ5ahy1DmqDbkLDP3q3XYgoqamp+jzaeTj9GuovkrvNGqWJ3QO/8AR0HXYnnPVOaoctQ5aZyUzkpnJTOSmclM5KZyUzkpnJTOSmclM5KZyUzkpnJTOSmTOgcnXX7f6DxfteivLwvVdVhSJV20jbSNtI20jSkaUjSkaUjSkaUjSiaUTSiJgoMuVoytO2YQcaeSLuRPUiOF9K67vx8mtUmsNhfnxaOkouPYmlE20jbSNtI20jbSNtI20jbSNtI20jbSNtI20jbSNtI20jbSNtI20iZK6J11+3+g8X7XqhYx648BjwGPAY8BjwGPAY8BjwGPAY8BjwGPAY8BwVxGUEKz4V/D/Uuu1zvJXDvNGruT9L80a7RZpGRK6xWe10VJTgrmPAY8BjwGPAY8BjwGPAY8BjwGPAY8BjwGPAY8BjwGPAY8BjwGPAY8BjwE0UbE66/b/QeL9r1QwrMuDIYMhgvMF5gvMF5gvMF5gvMJ5hPMJ5gvMJTCYUomRNmfsafNjfU1nmjHaK5Nrvm1F0WTTRXxPY6nEphOMF5gvMF5gvMF5gvMGQwZDBkMGQwZDBkMGQwZDBkMGQwZDBkMGQwZDBkMGQmruhTrr9v9B4v2vVHE+VcOwYdgw7Bh2DDsGHYMOwYdgwrBhWDCsGFYMKwJQsCUFKkTI2Su3yDHbXO/jkf6Xyof2MaqtV7dUavlJQ1XCsGHYMOwYdgw7Bh2DDsGHYMOwYdgw7Bh2DDsGHYMOwYdgw7Bh2DDsGHYMOwYdgw7Bh2CSCWJOuv2/wBB4v2vUzee4PcHuDSwaWDSwaWDSwbbBtsG2wbbBssCV7LiOptV68cH5+USeqKPzTza5ybka5WrNV3iwzoaToe4PcHuD3B7g9we4PcHuD3B7g9we4PcHuD3B7g9we4PcHuD3B7gfy6ddft/oPF+16o5HxrlTmVOZU5lTmVOZU5lTmVOZU5lTmVOZU5kWFGpckK0W11iTfJqampE7R+qxyyJtcv8jWvVqyxukHLYYuTOhlTmVOZU5lTmVOZU5lTmVOZU5lTmXOZc5lzmXOZc5lzmXOZc5lzmXOZc5lzmXOZc5JNJInXX7f6DxfteqKVYlzZTNlM2UzZTNlM2UzZTNlM2UzZTNlM2UzZxmTIK7hr7lNym5TcpuUcu+JjuRiPcxz9ZUdJMxUuToZspmymbKZspmymbKZspmymbKZspmymbIZshmyGbIZshmyGbIZshmyGbIZshmyGbIZshLO6VOuv2/wBB4v2vVE9rFyIzIjMiMyIzIjMiMyIzJYZLDJYZLDJaJbeRNe5Z5eR+pqampqRSbXPRY3qvMjnyRqlyQykUyIzIjMiMyIzIjMiMyIzIjMiMyIzIjMiMyIzIjMiMyIzIjMiMyIzIjMiMyIzIjMiMyIyWVr066/b/AEHi/a9USxou6qbqpuqm+qb6pvqm+qb6pvqm+qb6py10IpnPWR+1Otjke2TdE5La6c0Cm+qbqpuqm6qbqpuqm6qbqpuqm6qbqpuqm6qbqpuqm6qbqpuqm6qbqpuqm6qbqpuqm6qbqpKsKp11+3+g8X7XqiSNV2VTZVNlU2VTZVNtU21TbVNtU21TSobqrR1nVORpyNORpyNORpyNORpyNORo2zom6q40qGyqbKpsqmyqbKpsqmyqbKpsqmyqbKpsqmyqbKpsqmyqbKpsqmyqbKpsqmyqbKpsqmyqbKpKkKJ11+3+g8X7X/BV+3+g8X7X/BV+3+g8X7X/AAVft/oPF+1/wVft/oPF+1/wVft/oPF+1/wVft/oPF+1/wAFX7f6Dxftf8FX7f6Dxftf8FX7f/g//8QALhEAAgIBAgUBCAEFAAAAAAAAAAEREgIDYSExQVBRQBAgIjAyUnGBYhNCYHCA/9oACAEDAQE/Af8AVSH/AJEva+4z3CSSxYt27NxzNTKOEC1PCFq4s/qMw1uj7bXY1FLKkeUVxKmLntsEEM4kC4C7W+0twu5ZrgJR78FSpUqVKlSpUqR6/Pl7yFiVK7FSpQqLEWJXYoyhTYeA1Hrcvp92DHEWJEeyCr6oquiGYrmR5IRVeCmxTYqPGTLCPV5D+kRGxGxD8GGm+qFpxxyI/SEvtP2TV7CXhj/kOfJgpRicSdz9shbnDyUnqZafky0n0Q8WuhGxG3p3zH9Ps4HAVOpo5fbicuLN8h/yJa2NQ4NcRKq8EfCzTEyfwW3LfyE55M+I+H8HxI1NT7sDJ6fg+E4HD00of0iJ3J3J3Fn5yMNRf2lo/JMcuZPgyynqYZFkMwcCyL7l9yzLeUJ/ax6nky1MOjaHm+mRZ+Sdyd/UZ8vdkxzy8iy8E9CxYTLEiZYtuWZKJaHniPN9Mh55esz5e+s2WLlixYsXLFyxcsPN+uz5fIkkkkkkkkkknsGfL5D+Suw58vkv2yT7UmR2pj4IsSWJFuJdrsnwk1NWchanUs1+i8fgtDjIw5cH2xyUFiVKmOLXbalSpUjtqg4EolHAf/Cv/8QAJxEAAgEDAwMEAwEAAAAAAAAAAAERAhBQEiAhQVFhMEBxgQMxcID/2gAIAQIBAT8B/lVCjkrWQSl3ajH0Lrf8ixypXXa6SMYviy/d2pGsbT8FJBDIRXRjZE4NRqNRqHjZJJJJJx0kk3nFSSSSSSSSSSJ4dsbJRJJJKJ4JPo+iRV4Vjdue9+LP9D8nBwcEiqE5wdTHaL82duTk+z7smKrASVVE+g7QRtTFUT595ycjkqXd7VsfopdmKe5yc+8ga8FSfXYh3ezi3yJCpqI8EeCPfuldhruefR4OTgSqFT3QqVg6qUeSNsX+SOx8lNNPTCuRqCCBIggggjuRBpTEmuuHa2xsSxUXggggjGwQQRjYKfx8GjoaTQaeqKsZSrO9bWMRStlTQ35xlMdTUjUjUjUht/4V/8QAOhAAAQIDBQQIBgICAwEBAQAAAAECAzORESExMjQScpKhECJBUFFhcYETICMwQkMEomCxQFJigsHh/9oACAEBAAY/Av8AgN3/APA4e6ncLd//AAOHup3C3f8A8Dh7qdwt3/8AA4e6ncLd/wDwOHup3C3f/wCTalyeKm1c5vindMPdTuFu/wD8hG1NluRC2zq9p1crsO6Ie6ncLd//AJFv5RP9HxH+yGPsLsp1m32d0Q91O4W7/wDx0abKZWljcCxbXL5HxIS2p2oXZXXp3PD3U7hbv/8AHdF7VuQVe11xspmdiW2FuBtNxZ3PD3U7hbv/APGRqdo2E3BtxZ+MMWM/2QtVTZf1mlqLbCcK3sxTuaHup3C3f/43xXYrlHRV9iz8oimzgxmJY2G2zzOp1XeBY5Lu1BFZe5mHmnc0PdTuFu//AMXbjXJ4eJd7INgt7BzuxiWIJCbji4s+KlpfU2YvEX3sU+LB7S/uSHup3C3f+dfiqqehMfQmPoTH0Jj6Gd5neZ3md5neZohmiH7FPpQ7PM7XKf8AteQsRfYY3/6U2W5nXqXoWLe3wNqHehsre3wFYjuqvIsjQ0e3xsMr0MYhneTH0Jj6Ex9CY+hMfQmPoZ3md5neZ3md5neZ3md5neZ3md5neZ3md5MfQmPoTH0Jj6H0nOVfP7EPdTuFu/8AOu09GepqWmpaalpqW0NS2hqW0NSlDUpQ1KUNSlDUpQ1KUL4qu9CyEiNLsO1RrUyoOd43F2Z2JiWsud4F1ynXSxfFDahrb6FipZETsXtFR6OhuLo9UNS01LTUtNS01LTUtNS01LTUtoaltDUtoaltDUtoaltDUtoaltDUtoaltDUtoalpqWmpaalpqWmpaWtio/0+xD3U7hbv/OqNs9zFlTMypmZUzMqZmVMzKmeHUzw6meHUzw6kyHUmQ6k6HU68Xa8kNlljW+Bb2uwLPAtstcvIseli+PRZET3LW9ZDtQvvNrZtsx8T6cZv/wBGeHUzMqZmVM7KmeHUzw6kyHUmQ6kyHUmQ6kyHUmQ6meHUzw6meHUzw6meHUzw6mZlTMypmZUzMqZmVMzKmLKlrlb7L9iHup3C3f8AnsYlpLUlqS1JaktSWpLUlkslksyczJzL9lPcteu2vh2Fq9haWxFvXsLWLahsvvQtbehcYl7ULdmw+LBXZVTJb7ktSWSyWSyWS+ZL5kslkslkslkslkslkslktSWpLUlqWvbZ9iHup3C3f+f6e17H7eZ+3mft5n7eZ+3mft5n7eZhF5mEXmYReZhF5mEXmZYlCW8v6vqp8NF2lXMpaW9G0z3QuOuw/I/I/I2NnqqL8Pac3ssUwi8zCLzMIvMwi8zCLzMIvMwi8zCLzMIvMwi8zCLzMIvMwi8zCJzMIvMwi8z9vM/ZzP2cz9nM/bzP28z9vM/bzOvte/2Ie6ncLd/57YbrCYpMUmKTFJik1SapNUmqTVJqk1SapNcW/UU24irfciL0Wmy+vRelil1imBgZkOqh9Fyteh1lehMUmKTFJikxSYpMUmKTHExSapNUmqTVJqkxSYpMUmKTFJikxSYpZEdb9iHup3C3f+dVaiL6mSHwmSHwmWHwmWHwmWHwmWHwmWHwmWHwmWHwmWHwmWHwmWHwmDOEwZQuXkLEje1xapglha0sd049GBhaptM90NuB1mL2WYGDOEwZQ/Ch+FD8KH4UPwofhQwZQwZwmDOEwZwmDOEwZwmVnCZYfCZYfCZYfCZYfCZYfCZYfCZYfCZIfCIjkanon2Ie6ncLd/5124aP9TSsqaVlTSsqaVlTSsNKw0rDSsNKw0rDSsNKw0rC7+LDNmFBb7FsTMWuWwuXotTAxL0QylyIYnWLWr0XwGPb/wBrDSwzSwzSwzSwzSwzSwzSwzSwzSsNKw0rDSsNKyppWVNKyppWVNKyppWVNKyppWVNKyppWVOpBRnp9iHup3C3f+dfjNcqeRLi1JcWpLi1JcWpLi1JUSpKiVJUSpKiVJUSpKiVJMSpJfxF38dy+qlkODsM8ei3osXHo8DFC9UMTKWuXo8FLHFr/wCLai/khIfxEh/ESH8RJfxEl/ESX8RJiVJUSpKiVJUSpKiVJcWpLi1JcWpLi1JcWpLi1JcWpLi1JcWpLi1E+C1yL5/Yh7qdwt3/AJ1+K9Wp6Godwk9/CT38JPfwk9/CT38JPfwk5/CTn8JOfwk5/CTn8JNiUPzd6ltlieBavTahfj8uJaWKXdFim09jkt7WkyJQnRKE6JQnPoTn8JOfwk9/CT38JPfwk9/CT38JPfwk9/CT38JPfwk9/CT38JPfwmodwmodwifCiK5fT7EPdTuFu/8AOqLEaz1NVCNVCNVCNVCNVCNVCNVCNVCNVDNVDNVDNVDL/wCW32Q+ku25O1ei+9ejzLF6cE+S7AscXdF6FhtutZ6F38tlDVQzVQzVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQjVQhFSKx/p9iHup3C3f8AnVGWXH41PxqfjU/Gp+NT8OI/DiPw4j8OI/DiPw4jFnEfVjN9GljG2IXF5h0efR4GY8S4sLjz6cEPA8UNpOp6HViw19zFnEfhxH41PxqfjU/Gp+NT8an41PxqfjU/Gp+NT8an41PxqfjURX2X+f2Ie6ncLd/5ViL6IK3s7CxjVUlOJTiU4lOJTiU4lOJTiU4lOJTjLs+qm07ru/10XF6ly9Pn8nWLizsLU6MpgYFyl95anUXyOrsuJSkpxKcSnEpxKcSnEpxKcSnEpxKcSnEpxKcSnFr2K31+xD3U7hbv/L8NiLchYiLtIfTVyehmiVM0SqmaLzM0XmZovMzReZmi8zNE5maLzMYvMu+LzLY0RzU9byxC/wCTzLF+ezpxMTExLlL0LWucxfFFOpEVyeTi/wCNzM0XmZovMzReZmi8zNF5maJzM0SpmiVM0SpmiVM8SpniVUzxKqZ4lVPqK5fX7EPdTuFu/wDKu6L6lsNbCZyJnImciZyJnImciZyJnImciZyOqrvU+pEV7ui+/pv6LejDow6bCxS75/A6r1Yvih1Ym2nkdZXJ6tJnImciZyJnImciZyJnImciZyJnImciZyJnIsiOts+xD3U7hbv/ACrui+oqo1rt5CTB4STB4STB4SVC4SVC4SVC4SVC4SVC4SXC4SxjGW+TS2PY1PDouwPEuTouLF6PA7Oi5SwuLlPP5MDAw6LzwLls8zq7D08kOvCZ7tJULhJULhJULhJMHhJMHhJMHhJMHhJMHhJMHhJMHhJMHhJMHhERWMbup9iHup3C3f8AlXdF9RfiQ9s0qcRpU4jSpxGkbxGkbU0jamkbxGkbUu/isPpwGtb49F9Oi9TA8OjaQ8zDpwNntFLTaT7N7S5bC9PdC6EyI06/8RhpG1NI2ppG1NI2ppG1NI2ppG8RpU4jSpxGlTiNKnEJ8ODse/2Ie6ncLd/5V3RfUX421Z5GWKZYplimWKZIpkimSKSolSyH/GVVqdZL/wDqXVLr18TBS9yGctxNlSxTy6L0tLESzo/9Ym0WOTHovaYKZTIYJ8mVDBUOo46yKnmhL+Kh1v4yotCVEqZIpkimSKZYplimEUwimEUwiifB27f/AF9iHup3C3f6dqy4RVS5cBd0X1F+K/YT0NT/AFNQvCaheE1C8JqF4TULwk59D83epbDb8NvoeX+zrrsp4HUQxU8DOhaxyKbSe6H/AKTovbQufZ6ljMfE2lxQ2rfUtQ2XY9hsvOq9DM0viIZlO0ymQyIZDBTMqHVcinWRWKXtSInodZjmqTIiexPdwmodwmoXhNQvCaheE1C8JqF4TU/1PpxdtfT7EPdTuFu/0JaPXFqs6ou32u6ou6L6ipttb6k+FU1EKpqIVTUQqmohmpYdb+Rb6H0WbS+Kn/Z3JC693ieKn1F9kLGWNO1TrOah1YiWlj8f9lxjsqdVyKWWFi9Ze0tTLiLauYbb2Xmy72UsW56czAwUy9F72l8QzKfkfkYuM51YiH/ZDtYp1mtiIX2sX1Lv5NTUsNTDNRDqaiFU1EKpPhVJ8Kpb8RjvRfsQ91O4W7/Tsr/JRvkbXx9tRd0X1LGGVKmCVMEqfjUxZU60ViHWcrzBIbPBCxtzS3BPE2YVS4/7uPD0PD1JjTrXt8fAvxMC+42Ucps7fWHIuDRyqnmXmy/Ds8jZfc9O3xO0/Kh21L1SpfEaTSYtDOpM5F0VC6xS9FLlOtY4uVWKdZjYiH5MU6sdvuXPhr7n41PxqYJUwSpglS16fYh7qdwt3+i42liNaq4opakVrvJBd0X1Ooir6GR5kcS3GRTBE9z6kVE9D6TL/wDspat5tRaFmDfA2n3NNltzS1eq3xPpp7qdqnWc1C1jkcngXH0n3eFpZEZtJ5oWQmIy3wEV7tlSztXMOtyrgWeJsK1Eb2KfDi49jjqrankfkdtS9UqXxGk1CalCa0uc1fcyngZjrs90Pput8i9C5SyKxHFznMOpFYpgi+5LUluJbjI467VT1+xD3U7hbv8ATtpFZfjeW7bF9FF3RfU6iqnoZ3meJUzxKmaJU/YXNefUibCep1LXu8VL7zaieyF5tRPZD/8AC2LQsb1U8i1bvU6kRLS9LHdpjsu/2XLaXw0R3ibcRbEEbjbdtGz+OBcbGz7mxFw7HHUdankp+VT/APp2F72kxpmaYF6KXKp1kRxd1FMLU8uixesnmYuhr5KfTjI73MrqmWIfmZohmiVM8SpniVOu5y+v2Ie6ncLd/o2rHW249hhEPpo+3zF3RfUtYpm5GfkZ+Rn5EwmKXWr7F/VTzLl23eJ4qWrmPFS69/8Aro6t7vE8S96CI72cdZLv9H03nWbaWuuQ+mtqf6EYmCY9Gy7HxNiJe3sU+mtp21P/AOnYdhgZTwMTrsodRbS9C5TrpYvih9KMi+Sl7VL7ULnqTFM/Iz8jPyM/Iz8ix6/Yh7qdwt3+hE2kTrdpNh1NrbYvoou6L6irsNd6kiFQkQqEmFQkwqEmFQyQ6H40L7C5iInip4u8TzLEzdGyzA2nYGy25PA6zrC5dpvgeP8A+ly7J1XWnWwPpvW02W3+JcvmddVtU2XXtOo47a9OB2mZTrXn/Xpsd1kPpORF8HF8Ohghcp+NDJDoSoVCTCoSYVCTCoSIVCRCoWfDY30T7EPdTuFu/wBCM7dq3pXdF9RfiM2/c039jTf2NN/Y0v8AY0v9jS/2NK2p1f47EL2czqwrvG0sbj49Gy3A2nYdFiY+Ja5bDqOv6OvUueXuuLLV9SxmJelniWudZaWLe06runtMVMx1mnVUvLlL7lLWw9tPJS+HzLm8zrQWr6l/8VhpU4jS/wBjS/2NN/Y039jTf2NN/Y039j6cLYX1+xD3U7hbv/Ku6L6i/G2rPIxin7T9p+0/aftP2mWKdWAvuYOOo11niWNL8OjzLVLVMVPM8U8DwMxhaWMOstxaq2F96HVXpxL7FPAuv6L7y1l/kWPY9PYu2jrw9r1QkuT0MsUwin7T9p+0/aftMYpjFMYp9Hbt/wDX2Ie6ncLd/wCVd0X1F+JE2PY1ScJqk4TVf1NV/U1X9TVf1NV/U1X9S/8AkrwmZXepYjrE8EQtYt3iXYFvTYnR5HmXtO0uQuQxPI8i7pvQu6by1+HjYZ+RY5yOTzQ7W+hd/IdQ1X9TVf1NV/U1X9TVf1NV/U1X9TVf1NUnCapOE1ScInw4237fYh7qdwt3/lXdF9RU2mt3lJ0HiJ0HiJ0HiJ0HiJ8HiJ8HiJ8HiJ8HiJ8HiOt/Jgp7lr47X+5YkRiJ6m0i2p0WdHr0XHkeJgXJ0Y9Fil3Rehd03FjntRfUmsqXuY1fFrjqfyme5d/Jg8RPg8RPg8RPg8RPg8RPg8ROg8ROg8ROg8ROg8ROg8ROg8ROg8Qi/EY7dX7EPdTuFu/8q7ovqWQ0tUl8yXzJfMl8yXzJfMl8yXzMnMycy9Gp7lsV22vgny29HgeZ4dN5j0WL0X9N/TZEai+Z9KL7OLtlfcyczJzJfMl8yXzJfMl8yXzJfMl8yXzJfMl8yXzLYjbLfsQ91O4W7/yrui+p9NHL6GWIZIhkiUMkShkiUMkShkiUMkShkiUMkShLfQyKnqbURdp3h2dFh6iIWqXoeXidapc4zGYxLi3tLF6fPptYuwvIuTa9FJb6Et9DJEoZIlDJEoZIlDJEMsQyxDLEMsQyxDLEMsQyxDLEPqI5PX7EPdTuFu/8q7ovqWscqehNcTXE1xNeTX1Jr6k19Sa+pNfUmvqTX1JkSp1Vie6nWer39PqW9hgdWh1aGFhiYmJh0Wp0Wp0bLrfVC2FFc9PU6yxUJj6k19Sa8mvqTX1Jr6k19Sa+pNfUmvqTX1Jria4muJria4muLHvVfX7EPdTuFu/8q7ovqKrO0xbQxbQxbQxbQxbQxShilDFKGKUMUoYpQzJQz8i5bG+KobNtvmep6Hkhch4KX3KXpaYGJinR5p0eXR9J+w46/wDoz8jPyMUoYpQxShi2hi2hi2hi2hi2hi2hi2hi2hi2hi2hi2hi2hi2hi2giPsu8vsQ91O4W7/yrui+oqrDa/1NNCoaaFQ00KhpoVDTQqGmhUNNCoaaFQ00KhpoVDTQqGmhULoEJPYytNuK1Gt/2W9F+CdFjyx2BdgXpYXX9Fi9FvRtQWo9vMyNMrS+BCX2NNCoaaFQ00KhpoVDTQqGmhUNNCoaaFQ00KhpoVDTQqGmhUNNCoaaFQ00KhpoVDTQqGmhCIkJjPT7EPdTuFu/8q7ovqL8ViuTyUkO4iQ7iNO7iNO7iNO7iNO7iNO7iNO7iNO7iNMvEaZeI0y8Rpv7HU/jInuWJCVV9RLUscvZ043ljixcDyLseix3RtFi4HWhXdi2kvmdf+OjjS/2NMvEadeI07uIkO4iQ7iJDuIkO4iQ7iJDuIkO4iQ7iJDuIkO4iQ7iJDuIkO4iQ7iJDuIkO4hPhQ1avmv2Ie6ncLd/5dqy24VRfjK5E8iZFoTItCZFoZ4tDPFM8UzxTPFM8UzxTNGM0YzRjCK4+nDVvsK5qOu8ejwLUW0s7TZNleixcejzQsXtNl7XFitcqeCoSnp6H7jPFM8WhMi0JkWhMi0JkWhMi0JkWhMi0JkWhMi0JkWhMi0JkWhMi0JkWhMi0JkWhMi0JkWgnwXPVfP7EPdTuFu/867cRGGqbQ1TaGqbQ1TaGqbQ1TaGqbQ1TaGqbQ1TaGqbQ1TaGqbQv/lJ7IXxHOPhwMEv6MyenRagjkNpDZX26PNOjrLZbgbMRUch1Yzm+xd/KThNU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2hqm0NU2gmxGR/2Ie6ncLd/51RqonqZ4fEZ4fEZ4fEZ4fEZ4fEZ4fEZ4fEZ4fEZ4fETIXETIXETIXETIXEXxoSf/AEX/AMlhstcjl7VQs7V6PNBW0FYWLh0f+kLTbtuNiIrVb6n0/wCQz3LosLiJkPiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiM8PiEVytX0X7EPdTuFu/89kNtpKUlKS1JZLJZLJZL5kvmS+ZL5kvmZUT3OvFanoWMw8fEVei0uNpO28R6YKWfkhabbfcVrktauKFsByKngpL5ktSWpKUlKSlJSkpSUpKUlKSlJSkpSUpKUlKSlJSkpSUpKUlKSlLYjbPsQ91O4W7/AM/09r2P28z9vM/bzP2cz9nMwiczCIYRDCIYRDCIYRDCIZXe6lsZ9v8A5Q8Lfk82lna02F9jzQ22+/RtQF/+TI8wiH7eZ+3mft5n7eZ+3mft5n7eZ+3mft5n7eZ+3mft5n7eZ+3mft5n7eZ+3mft5n7eZ+3mft5n7eZ+3mfU2/f7EPdTuFu/89rHWExSYpMUmKTFJikxSYpMUmKTFJimdxi/3L37bv8ARdgnyX4L0XYdhtJmTEtQ2oD7F/6nW20JjiYpMUmKTFJikxSYpMUmKTFJikxSYpMUmKTFJikxSYpMUmKTFJikxSx7rfsQ91O4W7/zqrUT3MGUMGUMGUMGUMGUMGUMG0PxofjQ/Gh+ND8aFypQteuy3zLsVw+ZH9qXKbC4pgeZbCVEd/1U6yWexmPxoYMoYMoYMoYMoYMoYMoYMoYMoYMoYMoZWUMrKGVlDKyhlZQysoZWUMrKGVlDKyhlZQysoZWUMrKFjkb7J9iHup3C3f8AnXaYj/U0zDTMNOw07DTsNOw07DTwzTwzTwzTwy7+PDLGQ2p6IWx6HkmHzX4Lif6LrEif7LHNsUvsX1L4EOhp2GnYaZhpmGmYaZhpmGmYaZhpmGmYaZhpmGmYaZhpmGmYaZhpmGmYaZhpmGmYaZhpmFjYSM9PsQ91O4W7/wA6/Faq+hLfUlvqS31JTqkp1SU6pKdUlOqSnVJTqkp1S6BzNmDDRpsNW1e1fsfDf7KWOb7ljk2k8y+BRSU6pLfUlvqS31Jb6kt9SW+pLfUlvqS31Jb6kt9SW+pLfUlvqS31Jb6kt9SW+pLfUlvqS31Jb6kt9SW+pLfU+k1UXz+xD3U7hbv/ADr8Vyp6E11Ca6hNdQmuoTXUJj6Ex9CY+hMfQmPoZ4hcxy+pYnVTwT7Vi9ZPBS9jk9DNEJrqE11Ca6hNdQmuoTXUJrqE11Ca6hNdQmuoTXUJrqE11Ca6hNdQmuoTXUJrqE11Ca6hNdQmuoTXUJrqE11D6T1VfP7EPdTuFu//AIHD3U7hbv8A+Bw91O4W7/8AgcPdTuFu/wD4HD3U7hbv/wCBw91O4W7/APgcPdTuFu//AIHD3U7hbv8A+Bw91O4W7/8AgcPdT/g//8QAKxAAAgECBAUEAgMBAAAAAAAAAAERIfExQVFhcZGhsfAQMFCBIOFAwdFg/9oACAEBAAE/If4HSOz/AOD8dp8D07s/+D8Np8D07s/+D8Np8D07s/8Ag/DafA9O7P8A4Pw2nwPTuz/krHAnHAIkeYyfifDafA9O7P8AkPSUWLaIVCnRSQ8qtsEUYiXVfEfDafA9O7P+SkQVmfuMZCNlCl98QHhtPgendn/HevzdeAiXoH0iiFFFsh1LnHgFS2V8aBTvh9+G0+B6d2f8efB/pO4DQ602iIsp3FDZNu5H3PjbNfD+G0+B6d2f8bFBaBQ8UOSHTakqeMTND0oZNcUQWErehF+Pk/DeG0+B6d2f8acVlJ/ZrAqcQ5Kf4BEdw6htym1VkTnOY3VjEFOEP/AOjh/C+G0+B6d2f8RJtwlLI62/OE9JtoC5FxcSETvgZxBUK/QKCRyIJFB5BF4wikSeNpZjGhGmsn8J4bT4Hp3Z/nBDRT8SlKVZywlhLCWEtRavQcEN5SxQam7EMZNQZi7DjJUOlE3gvYUS9IWRzYsLkdVZoWQzYsO0jwlm4zwBOo4NMjQcjY/r81KUpSrOWEsJYSwlhLCWEsJYSwlhLCWf8VKUpNqCsPY8Np8D07s/zWkIscxYSwlh9qEILWuMY1XbEgULnXNkvKUBqBmjcKjQ4IpusbRDCjNaOo04NbWZiR6L5RwjmxSGGwpAgMUhvyx0eSWEsJYSwlhLCWH015D/AH2oQhCEIQxYSwlhLCWEsIjsTeHsPDafA9O7P83sK0pqj3XrOc561rXb9yQ1lYcMrkYjbt+AhaFglMxVyC4QSoqmVo0JaROkh0NghuDByo055k4zZGAeclQabP4M5x1rve973Wta10/YJznOdb19wm4xPY8Np8D07s/zam3VWXRF0RdEXRF0RdEXRG+5o33NG+5o3nNHhQ8CGar3DxRfBA7yOvgUTZ1KtdBkLe+CzQig/pGJfZoYSzQkauKK8f8ARFrNsyPDEJ4MddhC6I3nNG85o3nNG45o3HNG45DcchuOaNxzRuOaNxzRuOaNxzRvuaN9zRvOaN9zRvuaN9zRdEXRF0RdEJzatwvY8Np8D07s/wA67Ezr/KSSRoCNH4CEItgsQkx5ww37FBkwKLNciyY2Q2m6qxsND0IVf0DqXgyo90V6CNY4Q/Kipk+eJRD/AIi2CyC2CzCzCzCzCzCzCzC1C0C0+pCNgI0FGgo0fjSSUWHlX7HhtPgendn+bow7o4LEixItSLEixI8lHko8lHko8lfgIi/9yR8BrDJi+UHlPLAZyMRDyBgVGTzaRVj9B66L7NccAtrNvMk1U1Didh7EdqvcN3u7u6OfJR5KPJR5KPKRYkWJFiRYkWJFiRYkKDKpyqex4bT4Hp3Z/myhNRTPvKVylKUpSlIe0EFMUW9igocloJMisFqTz2A9ks1oNoLBjsapqQS9XMidTckUYZxEAkot0MNLhzegOwOwOwOwOwOx/AQg973tQpSlKcUirJzhex4bT4Hp3Z/m2krWDYfixjGXIuZcy5lzLmXMuZfxYK8aiXNPJTACWMYImUjJGGYnv6GiLYwQQeKNNSnBM35DmUvI1MhCNc0N7c8KnP2BmZyi1CvZfy5lz9tjGMYxjGMRUzTxbH2PDafA9O7P86FCpP3qUpShylKUsbNJEZdFYpQdbCyG5IxQW0MGN3CIwHI2FRgiCrnwIMENKSS5Tpr6HE6pvdRhfxqCBa1olKWVKUpSlKUpVp1Zz7HhtPgendn+aCiVGpe7PnOc4j3ve9Bi4wFH5JBUkwakIgieI8kNUY/g0OF9iixRqMUXok3ghhxBG6zGzkWQqApoZ/D0YtOURq+mPYfwJ5an8SMb2shnOc5znOc5ObSn1U4x7HhtPgendn+cfIUznLoXQ1eYXovRei9F6/H3d0spMM6ZRBRVY5J9B0ywGnBTMy1UJOZPPkDyqcB00jHBqCbQ4YCROoPeInWKsBj1CVoytYl54Bu7gcfNM7oXQuhdC6F0LoXQuhdC6F0LoXQuBdCdmOIfD2PDafA9O7P8Upw9KrZJq4PKp5VPKp5V/M0RETdDVWICj1JFUkxuMebM0l7EtmRilVanKdRtGK0OINBDiDJXO4mqsR6hgRCqVBUzRvm8JbJ8A1IlFKtb64OR/k96C3/QvyqeVTyqeVTyqeVTyqeVTyqeVTyqeVTyqeVSg3aFEvY8Np8D07s/xoDS4fcDgHppVWkWIsZYyxljLGWL8SIicVVNQT2Q55B6YwiXQsIowEVYDJkPDMRzL6NCgaMvRMQsbkGlS5Y6AqChk9hZlTaHjZOoiNKJpghqiSrueDkYnTs4P6JLF6VYixFiLGWMsZYyxljLWWstZaxWSnhD2PDafA9O7P8AFUldaoqSNsSx+bDM3d6OX4X4X4X4X4XUXwXoTIn+whaJ1sKy3C+2Qx5SSiUcDYIbaGDYMmgkVEywfoNtuom/cem5G50U1KtsS2Kj0EB4F6NOsAku4cspXI54YFy4XwXgX4X4X4X8XcXcXcXcXV+Ciii0k2U37HhtPgendn+PXu6OrDrNtQ6HgQ8CHhQ8KHhQ8KHhQ8KG55De8g5o1qSSJoxYvJcBtzCq+xRikSyoS9WLLUhqshJUPsTqmKJMR9jhQOShKhJ54EM4YicDuJmx6EtUS1RwPQSTVS9MsZtv1YTyKF5hobnkNzyG55DwIeBDwIeBDwIeBDwIeBDwIeBBRgWlUXseG0+B6d2f49e7o6sPnGURJ6whofiDGMbS9FbgAxvG7KqqsbhaImgsDWctjaI1Eiizj0JukKKJwmNJsLPoiKFbcdNPsVDL3DWInDxI4YokSzjeqFOTI8E8ULiBWxDRUlmv2zkM4RfuspWhCEIQhCERGc4HseG0+B6d2f49e7o6sP5KtUTcQXcXcXcXgXyXyXx6LS+7ZRs/YUNUxb6itV5JkTlQjceBv2LXMYG+yREcRIsOneJiGWg3hKTPiUaZVZkqKo22geYNM9hLNlenum++ZxeZx/Tg8A1M3G2gCtvaTryFYJvdR+As5vkvkvkvAu4u4u4u4WyHTq1KfY8Np8D07s/x693R1Ywa0KG6ouqLqi/oviL4i6IizuIjsmaBkKXQHNvnyoblO8J0QisIRpch6Qo1Qoy3g9CUXTMWU4tgxYBsrwg5qjbD6YnZiK1cUHN2+itCMUjaeDw3M4J7ekx5+ZpobVfR4QeMGwuRJi4lx4QV/SxxdVBGcdlJlO28JMoXZF8RfEXVF1R46JtSbUm1GoqSr7Dw2nwPTuz9WynGaspfmWOvd0dWEtRKOqTec5dpdpdpcZeRrt87Gv8AsDEXq0SVfKbLMUp0Ef21jciTZDal0bkOJsbg0RocR9BHxWiBEytGaBEdGBpSTmD1MjGWo8TOHToYuEsBCYec0Sq20uh/YBv5hH2BKYu4IjychwfmbT0DXTmSYq4E4HGIiw3SRpfECwMR93Md2WwpG/DyNvQV8F3l2l2l2l+m85xDbpKqmPY8Np8D07s/SOkmo4XbSSwiBrqtQQZ17ujqwt5BUy8EPw4UIv5q/WdgQXihebMCnMkC2c9ZycS2GOqk4KBAk+gy6hZK6sOcOI46BXM3AdQ26B/t5n+RunajQUkqQgIRzU04yZCyBZCalJv6mK2dUVRiLifhyhMzCzklxZFYf7IcH6RCZ/0Tr9Cl8glwdcUJ+J+4FFlzDMcGN4p4qOi4EHgpcBpncfw8hMCwOtaTXETex4bT4Hp3Z+iUuEIZDnGYHpE04Ovd0dWGpKbSl1gsEsX0ntisxBTLpzkxF9FRCKr9gMasW0isnDixGCFnnZSSdWzW/oRTE4aCutQ1GrPxICTrLURpkkwZq3gI2htuNJSdzHf6R43RloxA1JQ5cWNCEmxhE0+Uyh0SbrapIGbKfCcRFg/sixT6Xoi0k2DfYn7mg8/w2YVjE5cGKOFRuh9PSxxJdYjPlsRt7RAzTvnt+ju0yxRLSSbijn2PDafA9O7P0mleKaEdfy1gySC1Dr3dHVh40+zFmLf6LJ/uMY8YK7t1bGzT7pPlsYAllnZSVDBCeXH1GQO0syRe4MHOykzPq1m7rIx6NK0fYc8ufANEcXFBOsyHiFVm4IwPhvgQyUeeyKF2EpgAjImqQhZm3BWc3zjQPEQYP7Nbl+jNEJenDMD4VKubGHs+NTR7HGL3KMZQ5D+XLgNVge6qY7G9UNdRcCOHCBq/0RD6C28VGqfsPDafA9O7P0eFayY0oXgPCzkWzM693R1YetvniLuy7Cf/AEE3+w2G5neXQxSNybI/2GTt2zM/IybBktCLksax+kJBVyUTR7IX0V1C1IdjbWkNdzDbXyCr4PGSVbECVybN5joXgw2GGKlQJt8DwRUNCprkgmNkjHekEZpP7Go14s2n7IftNJ/2NOCPgzBV9GCr7E+izdGaPcwKtxAwF1HQHdEEP/0JmPs4SHiuQynTzIP9GXYXwXYJyULV7HhtPgendn6JU02IXNDYI4UHXu6OrDUkm1DpJs+Q2XKbLkPMg35uSHjha1P2kVN9+kWuNXgiU7lgghTyrQaqKhCYRag1w82yaKrWZClq7ngiOibElErKGeerMNxQ/uBpYHdSS54c2RWepuEIOs3GadtKVhBUQ3xlgNKA0TInilNnD9DlxXMa/Wb3ObR/ZijkN82NW8amWJ1EeV7HiS4gZishbAnXPqRnaWpMVcRGBMSv0G75DzIeZDZchsuUQ1mk5wj2PDafA9O7P0cshu5ePSL5VThUOvd0dWMZCohZ9ZmZZ+lDHgsvJhwGYgrOASELFTWfGhuWxmy2L0Mze7EMNmepVlruHUDyEO2hGdYxEUpkpOQbtRtGZCOIwraTUrYHcKIoljvMZsaxTkSZSr0yNQ+DMdJtNTcCXJ8zY6jRmHjUhgXNE5QiXdG6boPrxWqEqdGyjDxDFaeDCVrVJGp1V9GCBqy4ibFx13ekMiz9bMxtAczF7HhtPgendn6MzaeR69e7o6sNaNFFRB43PG543PI55HNhziRh936nRpseWex92z0LoYo3B8CZBSNz1FpZDBai4ngiEsu8KSpRXPtZkSnlqtGMOErozP24205o3wEqlLYuBS0ahtRD7YQkcWMV9DN3qcciEZUJOriZI4o/r7GNFHpl5XcRQ3r/ANQ0hqe/6GJp8Yug+wTj+iTE2y5jyOedzzuedzxueNxLTpKuufY8Np8D07s/x693R1YxakKGP10Rr6SNfSRr6TzoedDg6RKzvtFRYetY2oiWiSKDCYuqI2Fm9SRy1BgGGSMGsfQh08tyrEIyYHqKhLWh6jZKiWa2RLqNJYIafYRCzHmLB3I1oiZCo4NH2gyG5DmxRz8z/YGZUtjEMkJakCIoTNUCW33CwxfoUx9RMT4x+xuyvtE2J40I19JGvpI19JGvpItSP10R+uhIZtKvsPDafA9O7P8AHr3dHVhFJVKjcpL8L8NZeY2XMbLmNlzHiY8ziVieDCMlt7XBGi6MGRE2GuMGh5olBUirkNrfYjSlgI2y0QtIb3EoWvaNOcY1ajKNCEpYP2S3SOTesDqMEwIpNSGuL9M3CNwm8Bm0PRmIIWQUwoepUBsM0eFjYCDWb04xjA8Z/wBG+5zyObPmNlzGy5jZcxsuY0l5i/C/C/BpIduqUI9jw2nwPTuz/Hr3dHVh6SMpmD8ec5zWBI0jSNIyV0SY4TRpFISVa4JKNFGS1mNsKEPs57MlYjJiWQx5SkeIKcKDcpsBkYeic4VRCboKDz1ISsyD/oIvGQNVHVEP6R1VmTRCT/hH86ACG8FxI1gJGkaRpG0vYHOc5znHMlOIk9jw2nwPTuz/AB693R1YdYspdTyIeRDyIeRDyIeRDyIeRDwobHkOvCUBdHCJUJKElgkKmXixZcvAX1iePKUSXVsT3BWfElMnLT0VGqoQTbwKgpIHLkMD7HnVV6Qo5yCuKJ3O9mkOJg8MTtROx5DY8p4EPAh5EPIh5EPIh5EPIh5EPIh5EPIgowbQqr2PDafA9O7P8evd0dWH5MMzZ2WllxFxFxFxFxFxF6F6EuPPD2m+oFqP4JgK7w8I/ZzhH0VZTWnBE2B9GUNzkGx9CATzTmRzQjUG2xDwGDEo5BDVVQmah9AaacMUKYp4pjZt6YwYTUsS5wvYvQvQuIuItLLWy1stbLOyzss7LOyzss7Fppsov2PDafA9O7P8evd0dWGRpdGxcS4lxLl7FKUpSEHmD7C0oGrEipeYtui4CSl4LEau8EKhvWOJbYKJE80g1VLMKdy1EtizEap4URqCZSjcqK0ymP4NCVdH0RSi7kNDx/RNjOOISQ/7RJBcPZoQhCEIS5lxLiXEuJcRUS1hL2PDafA9O7P8evd0dWGtqUhypLZLZLZLZLZLXLXLXLXLXLXLNHiukKe7KwTClDHUZeYwoskDSz/0JtDgJpjdXzIhhJqS+ZENnRscYh6BO2eJFVVYY0omadFpNMgklvGCL9BF+stctctctktktktktktktktktktktktktktkZ3BpUQ9jw2nwPTuz/Hr3dHViXkKIy/mYxjGNp/lpSkITGeFF0XdFOj6hyHrsEYZ1qQ8EszlRGg+xvUZGPGaRqBsSdXQqsSTSYm+GLDhgzDOq0EWwkOA0uHJ9iWoo7yN6v4XRCFKUpScYxjGMY1iIGY5lcfY8Np8D07s/x693R1YYUSooPctZCEIQjs53mQZ+cdy21qYnSxMEg8bPspj0cE6Ia1V2CZPLJnjsOWlWI0Ogq2jEus1EcTKBPsTGcIYxvg0MSNRjjvat17EuPI5cP8P4rWta1rWta1rWtZJT6rk9jw2nwPTuz/GVxUIk3kclKtUjmWwWwWwWcW1FtRbUW1FtRbUWpFqRakM+vFQJY3yE9xjEM6ZiUKFjxDykKkwYD1GKwFbPYWhyEi/QY6OGSIvYDfM5NKjK5w4oDOW351IfNORaUWUWwWwWwWwWwWwWwWwWwWwWwWwWwWwWwWwWwWwVy1WPseG0+B6d2f5pZKli1M+9a1rWtXve971Lym6f5rf0QepsRKglGpkE3eIKw3VD1wTIr+4RIVzDwHRolSr/AGKmKtNVQaLxo5QznZHIn7pv417Wta1rWta1rWta1rWYzNvBKI9jw2nwPTuz/NlCSmuPfAZznOc50KUJ0UOL2uMLfq50KCYTuNqb96HWaqodOHjVCH46jKBPAcTLc3I5GQwInx2OtuiHEwtHNcfxynOc5xgAAAAAAAAAAABJVnGN7HhtPgendn+bow6UsviL4i/I3/NG/wCaN/zRv+aN/wA0bvkN3yG75Dd8hvuQYx4xBbldRA95mw0wwXpE5MxKuDFcBYcGgIb/AGEXRO6ElMUIyl4NBKwBWDHMrOhoacz4IX5F+RfEXxF8RfEXxF8RfEXxF8RfEXxF8RfEXxF8RfEXxF8RfEXxF8QoMq3Cr7HhtPgendn+dd4mdZGkI0hGgNgtj0peazzWeCyxssbLGxNw65lx9BGSWriU1KVKSy/B41grtwuAyc7HqE9MC+WNCcKqeK1Gs8fi7w4DOvIZpn6ZGgI0hGkI0hGkI0hGkI0hGkI0hGkI0hGkI0hGkI0hGkI0hGkI0hGkI0hGgKDAyr9jw2nwPTuz/Nqad0ce7/8A/wD7u7u6wb+BlgtXhDoHZryEymyUS1JaktTFmQxYk4XYZO1VWOXBqTc+iqHMd9jEQF/ov5P7/wD/AP8A/szMzMzMzMyc2qqp9jw2nwPTuz/N/C2oon2ClKUpR7J7J7J7J7Jy0+gX472sdBMlNH/X5AA9QJIOznJoqJkJqMMzgRZj7jJD4o10/wDJlKUpSlKUqxjGMYxjGMYxr6JJzgex4bT4Hp3Z/m9EjWGQvn6/PLu7uzFmLMWYjwH0OuVPIlB1cUW0dL8n6RpxF1xYRPEZPD0ZMmoReyTNb1Xd8/RfP0Xz9F8/RfP0Xz9F8/RfP0Xz9F8/RfP0Xz9F8/RfP0Xz9F8/RfP0Xz9F8/RfP0Xz9F8/RfP0K7Enj7Dw2nwPTuz/ADggYpL3Oc4pSlKUcuseIgxM2lgIbJPv9iTsNfWVpLRMGJwOhWPB/I/lk5znOc5znOc5znOc5znOc4k1dWXseG0+B6d2f5wQMUh+alKUrnnP8POf4ec/z05FmfQscu8TR9EN58jefI3nyN58jefI3nyN58jefI3nyE0fRB45d5NlfX8tSlKUpSlKUpSlKUpSlKUpSkmpqpPY8Np8D07s/wDg/DafA9O7P/g/DafA9O7P/g/DafA9O7P/AIPw2nwPTuz/AOD8Np8D0rs/+D8Np8D0js/+D8dp8D0js/8Ag/HafA9I7P8A4Px2n8H/2gAMAwEAAgADAAAAEAAAAAAAACAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPKAAAAAAAAAAAAAAAAPPPPPPPPPPPPOPvPPPPPPPPPPPPPPPPPPKAAAAAAAAAAAAAAAAPPPPPPPPPPPPIK/PPPPPPPPPPPPPPPPPPKAAAAAAAAAAAAAAAAPPPPPPPPPPPLT/3fPPPPPPPPPPPPPPPPPKAAAAAAAAAAAAAAAAPPPPPPPPPPPHjGQNPPPPPPPPPPPPPPPPPKAAAAAAAAAAAAAAAAPPPOPvvvvuiirXDhNPvvvvvvvvvvvvvPPKAAAAAAAAAAAAAAAAPPPLCAADDAgGZ2MAiQADgggjDCAABDPPPKAAAAAAAAAAAAAAAAPPPPIwx/wD98VRGNkDV3MHHHWsf/PNPvzzygAAAAAAAAAAAAAAADzzygwxzzwxAwrsHR+BDLLLBzzwwwz/zzygAAAAAAAAAAAAAAADzzyj33nHHGBT8tXhMGTTjEHHX33333zzygAAAAAAAAAAAAAAADzzyw8sMMMPLUVpVFYqdn888888889nzzygAAAAAAAAAAAAAAADzzmk884oJOJgJY17xaKX4MM8888sNLzzygAAAAAAAAAAAAAAADzxrAI4455cYsccK9oEgAEojKYIILLLzzygAAAAAAAAAAAAAAADzypT2kEWVz7iHt2bTKd+0o8InnHX2nzzygAAAAAAAAAAAAAAADz2MBGEcymtki4NeEYcUd89pqgssENDzzygAAAAAAAAAAAAAAADzqNDrZTppLNzeLwhmx7TzzgrE7F7r3zzygAAAAAAAAAAAAAAADxepRiUAwhkX2tE8fS9YFHn9krHhhDLzzygAAAAAAAAAAAAAAADwoMS74kv92FoOCQP8nMrvmEHxrA75nzzygAAAAAAAAAAAAAAADx2oA89MCCodR6Ka+/v8j6H7wbLc88zzzygAAAAAAAAAAAAAAADzypQ1332i+F8CLZQhqf14PmXz3y01nzzygAAAAAAAAAAAAAAADzyoXopKJroIUThRPX0cvxBqJbLb767zzygAAAAAAAAAAAAAAADzyoE8889/sMjfvsxwG/pi9/8A/wD7zzzvPPKAAAAAAAAAAAAAAAAPPKgQzz3882miWi1j7ngf8884wwww1fPPKAAAAAAAAAAAAAAAAPPPeMYQRTTTDsdMM/5nIUccccccccdfPPKAAAAAAAAAAAAAAAAPPPKPuMMNvlIs3ER2MzPPPvvvvvvvv/PPKAAAAAAAAAAAAAAAAPPPOM41z/8A+2Nzyx7MPPPPPPPPPPPOfzzygAAAAAAAAAAAAAAADzzzgAADDDBbLLILI+AAAABDDDDDDDDzzygAAAAAAAAAAAAAAADzzyj33332Mb//AP8Aev8A3333333333333zzygAAAAAAAAAAAAAAADzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzygAAAAAAAAAAAAAAADzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzzygAAAAAAAAAAAAAAADDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDDCAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAD/xAAsEQACAQMDAQgCAwEBAAAAAAAAAREhMWFBUYHwEEBQcZGhsdEgwTDh8XCA/9oACAEDAQE/EP8AlSxUXXxBKX2tR4euvauvhyS1JRKIEoheGyMuK7EtDIlPwxyXjcDYJJkqbqK6MAzMvUQnIqk+CxgjBGCMEYIwRgaaIXJwNne+jHNli69BMqpJDkqqkYIwRgjBGCMEYIwRgjBGBLHgD3geqCERoJFiB6ypQZMggggjwVs2EVoRqQiERSxAnAn4CuZicqV+bII3IIIIyNEZEhPwBTdvQUkL8kpJkyRIkTJEjzdjzDdEd+uiOTk5ORJdzRIW+CAjwyDo6MbLYaLIiWkbKJ7DAiWyGwkJFzk5OTk57u2lceWgRDIZDwJnsPS+idUU9bje5wNRVP1RG8Ud8MgpIeYCVtDFNNFJIqzkyv3MjrzIdV2LxfyWS64dRjXRDwQ8EMhj7vuPsSEgnWC1U8/QhOjrBJqW/k/o2XL6kcWfpQkq3ddITH94mjek+4qNymi6+SYbVfgZpQmvJ/p7DaXaXJV9V/pCf0f0ZPceeusobLRHx9DIo/f9j1UPyZdBISDTi3dnhKwPLVkXabRsSqR5HDXXovr7HRqo9F18n+C6r8DQq0hVVOfZ9ZE7I/OnuIgShirZShLrfrYjDTT40yQK65s/LbyILVDrS/0hu9F9DNT6ErdJ+xW6qXobFfk0phizrkeyew9ndt5jptHYkJBOtQ7U9x1XN7v7Km7degpoq3dfAlT1PVsQmrj4YxKJ/fsJFU0s/wBDJyr+3LJDTbXkQ1mPh/QsHXJ5fXItFp9eY0f+wiJ+ggVP7GgNQe42XYkJBttX7slBdESyWSxMrQapUuPoTrl7m0/MaNY+RtqyF3Jb+w578kLoxQrIn0Q3IZPejN5JIluGdc+hi7Xt9EvBLwSyWPu90RwcHBwJxoQIoi2CVydiRPTsSmScj3JC2E7MmrDN0hudDg4ODg47xdFb83D+IAJH4BdFb80ehPYmSSS0STFGI3VMSevgF00/NrAjWkIdiew9xMVJ0NQhJt4C0mof8Kq9ZHBJf1HHj4KKqxFO9GK7IIVvBJyTknJOSck5JyNiEmT0WRIk6wxqOlfgNHtV16oS6E9f2KlconJOSck5JyTknJOSck5JyJ57+5KDJENGSuU0di0OnhkSeQ8p5TyDVaEJeGQXZIyGQb3DPf8A8K//xAApEQACAgEDBAEEAgMAAAAAAAAAAREhYTFBURBAcYFQIJGx8KHRMGCA/9oACAECAQE/EP8AeEpGmtfkVpIe1MfIVAq0GpqSZHx8a6Ki0bvxEEEEEEG8G1C1L2saTVCJ16IIIIIIIIIIIII7hEYIwRgjBGCMDtOgm2r1LDfhlb0PQZwRgjBGCMEYIwRgjBGCMEYIwPx3NFFFFFCgfZewsy24HsbrkaMmD+CRSkyiiiiiiiiiiiiu5nJOSck5JyTknIn5ZGI+jBk+Rz1cmqick5JyTknJOSck5JyTknJPwE/WhPxTY2T9AknoT+IkYcjIl7MkS4NA+R5CHKT36GWWWWWOXTe9YkiVoNNGLgzJIqIJbEsSwmWzJhOdCyyyyy+4WvT0RgjB6GhaG7oOXf5J2CXrvuWlVoTbYU7KGaUogPI9hPIWf3IrggTBGCMEYPR6F47dajKKKG1kQ3/YxCe0ltiH7RC1TgyYlXKJ1dl7EcDyJYlhevwQORT5JWSUUULtkbjJyTkaLc2ExypELyx5ZGBqVkeUJcCS4NY1JA8CMIjwQ+CUiDRi92xKxK5JyJ57da/QZUb0hroYQnwfz0WnQ3L5JtdLPuRgjB5RRZKYsE4mLoKe3WoyFwQuCHA3ZBLUjXwTNvQnkpt0XkTixpGRhkYJCwxvgSOmKbJi5IQEOCMCWO3WoyiihpPk2jsYntJ3ei08/gUD0K/WJ/qHLWhgaS1TJ5GDklsDWilChR0TIWSslFC7dajPZ7PZ7Gp3Fbtshq2pSvuNbsSr9/f6GufwL9o0RBErT+xJbOByuRewlLqzVM0JRuez2ez2exee3WvS+llliNoyVMyxRoZydH7vpfNjBPodCc1jpiyyyyyy+4Wv0R1gZsQQQR0RggjZjdhLlEdIII7ta/4I6II+oEL4pISn6P4jS+FjBGCMEYIwRgjAnSlojtqOLt4Er3piktbQ5rcJTaIwRgjBGCMEYIwRgjBGCMEdzRRRRRQoEtpQSaJMEze41qrKKKKKKKKKKKKK7hE5PZ7PZOScl3qQKX1jYS1pjBOSck5JyTknJOSck5JyTknI/PcIsssvopwMisjOZzONu5zkyyyyyyyyyyyyyx/8D//EACsQAAIBAgUEAQUBAQEBAAAAAAABESHwMUFRYdFxgcHxkTBQobHhECBgQP/aAAgBAQABPxD/AMm/sGj7FPkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkktGj/ANJPtGj/ANJPtGj/ANJPtGj71PwCubj+hiV2Inu+02jR95nvy+EGLIprSo9yUk+GM0ClWVNotO32i0aPvM+ZajoWqS/yKsni82r2HQ0bCjQTKKMmTzjZ1+0WjR94nzHRWaZn8FNhUEsFiD6JDkjMfxZSAkexSGM3C2NMx7qoZFhDxXZ+Ps9o0feJ8H4/su9DQLd3VfQwoIhY5CvcXm02Ekp+RVxGuCabjhXTVMVY+F9ntGj7vPUNKyGETJN3m73GknW6U/v6MLcGZsvgaHDeCTouhMVRNVbiYr5cGajVypJq+Hxh2+zWjR93nrgxLe3xt5MxUNzzfMy7FOdvIp0obuff+kYIpMncJR0NWynQxox0v3iZpjfPUb0/G4jYhppw08vsto0fdZ6EwyiSVWJoFq3G3WSI5idJggh4l8Os8z7VFF0wfXB/CO1onukvV3kYZ+6dRkJCxZo+jFAQ0UqrqYzBqJrVEBGGVb1v+xnfaGiGvslo0faJ6qW6qy2/gvHBeOC8cF44I86+x6Pwej8Ho/B6PwRZ3ZwekcCT5b2HCXgQw79gYhzXCSk/4hMEoMpVLyQGUubNsXe45jq5Y9sPzHwIk+WoSy+BvwjFL4G1bph0Go8rOnTZDoVD4OcbbYTizhYkbvIfyceqS+WS+EJMNg14LxwXjgvHBeOC8cF44N+vsej8Ho/B6Pwej8Ho/B6Pwej8Ho/B6Pwej8Ho/B6PwT5V9i8cF44LxwXjgYeCiJJfH0LRo+0T0F6RIpsxLXyWvktfJcOS4clw5PYOT2jk9o5PceT3Hk955GNHbkOTUo8feGvTictd9Rz3VJLVLFlrRt1qJmhqgcchE5zI2DHtVNWOm5E20pjXKIkNd0H1QlYmJBP4HdSvJAq7JX0T6Mxd2sfBJWBvHyWvktfJa+S18lr5LXyWvkix7H9f5BcOS4clw5LhyXDkuHJcOS4cmnLv/Ra+S18lr5LXyWvksfJFbg0VS1x+haNH2ifOnUpoL5waiuv8Fm4LNwWbg1L3Qt/Bb+C38Fv4NVF9jWRfYTL8ZMWHlaCT7zwIJlIlXqRgcUCx3XsQUaFurq/AqB5MlZCEkFuRLuIaRirmt0xJRLRJr3F5Oox/BW4TZpjjrsplR3JxFoSnfWo9OFrSXf8AhNhmq/gs3BZuCxcFn4LPwaqL7Fl4LLwWXgsvBosvsWfgs/BZ+Cz8Fn4J8V0/ks3BZuCzcFm4LNwaTFbQvnBDaUrJ+haNH2ievWcCaULuXx5L48l8eS+PJfHkvjyXx5L88l+eS/vInW35E7PZuJ2VdeYae6Hxoxa+QhHvrdDLHSWSLBGF6o3XW/wJsgrZqbjI8rECujIefSSakwxF1RIi6pwLIaJsH83CJg86ya6GOLakO8MgyRNXE/yXx5Lu8l3eS7vImW35Li8lgeSwPJYXksLyWF5LC8lheSwvJf3kvzyWZ5Le8lveS3vJfHkvjyXx5L48k3Yw2qvt9C0aPtE/ZDqo3g6fgOn4Dp+A6fgOn4D0I9CPQD0A9AJM4G/FxPHPqLHO5t/YSpqrXikgPuhhCyWwuCmBa6CQeLDcMl8lYssUOHROLbEzLZrJmdLpsip2KHqkapIypUYl0cG83qPlpbbMgtGp/Q3IdAHm/wCZIm4GS4t/7Pvrrrho8f8AQYM9QPQBpx+IepHqR6kQmXwHT8B0/ANpjAO+Z9BO0/QtGj7RPQMORE5XctzwW54L88FueC3PB6Tgek4HpOB6Tgek4Dd/DguVwXK4GcSN5KP6FD1vEv0Te30JnASUpqtNXkTvb1aCyhN52KY2GBzWDQ25mcVUY8ZXs6lXmhMwLVoQOj1hktKzBVE1w/IDDrrVR+YI3x8E2HxLgk41wX64L9cF+uC/XBfrg9cuC3XBDwLg9JwPScD0nA9JwI+PgW54Lc8FueC3PBbngtzwJfm4CN6YaKH2X0LRo+0T5qsihKuRcOS4cmivp/R6FyehcnoXJ6FyehcnoXJ6FyehcnonJpL7ORIq61SyiN6AU/wU+DqZIzcDY3pa7UkcNbJXk+8Ca+H2fUbVXoVQlTlSu5Q8oeA23cT4irqhrp2ObBD1VGrlsQaNLgYzUYtD0nTFOogwT0U9fPXz189fPXz189LNRvYerHqx6serHqRqs7D0Lk9C5PQuT0Lk9C5NV/ZyXDkjIiOU+v0LRo+0T46kFETTie7cHu3B7twe7cGn8vg9x4PceD3Hg9x4PceD3Hg9x4PauBtLDVOAqlwmF86CruiSr9hCZ03qQLaq5WSafdZNDebYrUUt0aOqM5mqUEOZ3cmZqZIb25q7iQtbATIlB5GKzRhYb9iKuGmO0aaGPApPGn8NOLp/D1/8PX/w1Z9vRY/w9z/Bt/QbXg9g4PaOD3Hg1uw/B7twe7cHu3B7twe7cHu3B7twe7cHu3B7twT3YNxtNPoWjR9onuZxooTTnM9Z5PWeT1nk9Z5PWeSHM3acnr3J69yevcnr3J69yQYM6/2XLknAOuviSSyASRvRQqsX09jUxoozrQRylNVUdIdGMZ7DEuzMx17NJqUVY/uQ/BHImYaaJVkiVYwbG6wcqg0FcwTBimTT3FgxKNTLKMCNdWHWMGYixsXKV0qZHclP7PZuS7cl25LlySYq3X9nr3J69yevcjfk7JeT1nk9Z5PWeT1nk9Z5PWeT1nk9Z5PWeT1nkVS7VWjUZV+haNH2ieoipMnbnCCLFnRuD2Hg9h4PYeD2Hg9h4GrN6NwXjgvHBeOC9cF64E9OzKHgdkpOpQ30WIhVEjodI1FRn4fcayN/oZ6tiwtJMr/ITUYk4ahhDluQNIY+iIEZatrAU4lFiksSOEsBGxw5GZjECC5vMTqh8QSdMRM02SHF9MEY2dhT4FGKmvqXXgufBcuC5cE+T1bg9w4PYeD2Hg9h4PYeD2Hg9h4PYeD2Hg9h4PYeDSR1bg9w4HkVBlFr9C0aPtE9LBiPhVYL/ZzhwQtHJaOS0clo5LRyanZf9Pef0vP6Xn9HCpep5GLGaq7VEIYqXLgvIjVXxWSHNSEtEhRQ0y2sGPbUaIOaanBoUwpLR1JcR1DKiC7BJNnOWojMJm2sisYZgQgdMM6MUy1kxPUaxkxYSzVKGNlay5Fc43BvVqHH4GjW0kX5kydwsvmSX9ink1ZHh/8AAnOc5znOc5znubGLMlPc/oWjR9mnsaEbe3+YxCI0iYzNkWyLZFsi1o9pv/Ab/wABv/Ab/wABv/Aay+qCy21acCLKhVFntL4EBkYLN7t5sZ1tsOZ9IKme8THIny4IPEyvgTcJMWIrSzTFEn7FCKTk2ohKtsPnkFKWMRJFqmoGjQTitTdAk6uqJPFdhOVc1YY3ZhiniNnlq0KvfJ2gSfTU5vJ6wekGhPopsi2RbItkWyLZFsi2RbItkWyLZFsi2RKowQqu30LRo+zT1sraJoaxnF8U7sfDOqTVsOOwqG0DLS/7EAAAATRetR61HrUesQuLPCSj4liw9UoVttXuTMnBxeuyGtPTp5I4m0xKU7SsRjl10RiG2mLajVRlnJ4MxJ30qNNOGmmSyEtE2JRqhLDUYUKSTyKQnQa1FQd5BJRBdGjNqdf86XHomKaWtuIIQzcK26hrlnvwGjeW3O+HyNIbux/pifwj0i+spSgAAAAIhCF0M4aCX9C0aPss9Y1MZ05Uk6pPB4tsdCXAUNZrEbhKhmacbx/2aYkkklJj/hXJ+DP8jFWa96BOFkufYYwhCwuFLbOreLKi7KjwqaISVLkGmw0muFmbmULJ6DIE7mJC7igxT7EiaWTTbosw5khqiEHZcmqRqhht8NGYW2Myfwbz4NwJsE7oqTnRiYrJ6EL2zc1LrL9DNvKzWfw2P4pM5ZDW4aH/AMxSSWUWP/PGmmmpWH/C6aaIQ1yjiT2n6Fo0fZ56C6alTWxByu5d4ze37G4t2Nxbsbi3Y3FuxuLdjcW7DZY/gTXCa9uAR085N92v0N2LISWxyHoP6Q02WaKHDEV2K/mHPNNcUOQyVimKyt4NGLpRoM0T0ZGKegjh0I6klyQ73LNJQbxjS5vNkJqcfok6pOpaZrQu57E9maTboxLWa+TFLUXBtMMTT6pjRNsP2hfqRhJ1oIXj4LR8Fo+Dc27G9v2LvGXeMu8Zd4y7xl3jLvGXeMo70IofZfQtGj7PPQXTUb0MIKrity68l15IcA76lz5LnyXPkufJDgXfUuPIpyzBht/kcsJ5erRGapGL06HzxPMnE9DAN1+WqaxWkDk0FxMd0JxaMYmS0aeY6iVo1Qk/GsYdKVi2VFA+Rgy2qvqM4huvQdDaMIgo8pLgGfDF2N8T8FPYSbD4DdfBuBPAT2MEhNAccpvaqhzcmqLh3RDB8H+Qz/UjvpK9P8st/JZ+STFqvqXXkuvJdeS68l15LryXXkuvJdeRFQrJb6/QtGj7PPQXTUgKMayTjQsjwWR4LI8EGEG/AWl4LS8FoeCDDuNwMU143PA7XQs18V7CUNNBKxbbYrirNFOojy133CG8yaVkeHVMoK0U1QgtpqkdGYGzwaMV4VYN4DEF7qpA6xZLp1OiEYoq0TTkSaoa6HkTx47SWooOnVOfkhpTYM2KFM9T0Y36LsNub5JOQb3yCTn8hJxh9UZy7qg45bpYRCs3UldzV4OSG7L9SMqzxUn9E2PZfg1vl8HvfBaXgtLwWl4JMexwFkeCyPBZHgsjwSleNYIwr9C0aPs89BdNR11itTM5yR5lmpfHkvjyQZlmpc3kvbyQZ92om/1XyJMZeJG/LH6xylI1HXFBsxhRnwIZE3M1NjEvqKIz56TLJ6KTcQ1WuyVk5lmENqdE21IhE6pOBhjWkiZISborKZiqmxVGKaIwY7KaQXgXPXBErzHsS4RPyTebrNkhjs2NyHN+DIN6iPC9ak1+AGikuw6HwNR3UfqQoPy8aRF0fjFTdlXeCQoPMF2Y+y4xTSXaUN6vRrySZF2pe3kubyXx5L48kuR3Hvx78e/Dms1KI2j6Fo0fZJ6gop1cTaSmixGt0si5TT/xBdNRkFEkNNOFP80uzwXZ4Ls8FmeBLxetuISfwkAj022soF+EjVotd0Sq9RC0pXPuO0rrwcfgwKE9OWT4y8h/pEEVqPA6BZ7VKMBijx7DGw7RFVhN2nKK0qDGKjcVSjWTVoFUdAQVALFX6Ey8gY8oETVmQkLAyaCq2xlsnS2VjOwkalGFX2MW2tHhoowRyx6JP2lV4Mh19DMLqxHh33/Cy/hL+kx+86RYiOsKmesj8FXsy3IVRe7UdhYWYpVjsyv7Q5CfBfVuCbBHXiJ8EdeAvzwX54Ls8EmHe/w8abEGNFrX6Fo0fY57q5JVOyR9YRqZEhIdkXKNqMXGS/xBdNTLQ0Z7L/OQ4/4h6q0dWO+okYX5eRnKNil+5Fby2FGC/A7VYyeH9PyQdv2C2SFGruQjJF45K9ztRKlvuTCRq3KXcgTtlUg26mGAglbDRzCLAWLvNaDSY8VEsyKzt6gxOWA2qp8jRIaxToGaRS9dERoZIbiXwNcjQmrX9GjfKUU2DHBZgQfR16Ur8DSo6OYT3VH7GlKPZOSMzo7hmOgh4Is34LrR/Av8JrCX2UyFCa0Dal5E/ufB3RMsqQNCCOOqXTs5PjrS+Bv/AHqeTTs9f8tof8cXuZ7wiLd/QtGj7HPclEtuEhLVqmgz6kTqaSmTj5/xBdNSs9UJR3L08idinqHogs9XVD9lf4MObqvA1T0PzP8Af5FPSS4ONxOGd1JFDbBdBmyEwGGbsS6sjRGWcbRdLgR1pmx/SKi2hSFmo02Zlk+h2JhU+B9jTV5khaBohiFhtCydRVhb6Kh6FeHtxpOg71r3mq9pFSKpOknJJVj8DKiUVDbIh2Qh2BlJQksHRMLHbHwc5h5HUbgX4EKRnL3SPFnRuC68GWXUM3enkNKUh4YmsU+RzPZwd2Xqk6BbcjaNlK3YblvpQXZyyutrJcwawzZuB5b+g0mdhot7P8MeW+zUjm+Q1dvoWjR9jnshJVGm47r8q+KtRtmZKqfX/EF01E1EpRW2l2PfCdecJ3ALyZF+2YHuq/EjKHLG6/Aq/NPYsvwTknUT2MxjULaJdRG/Q8dqKSPRY3URLpq83ohrR04rlvoijS/TJeERK7pIxq+vWLoKDY1NZy1YlJIzNi0NuZSUnyhvWoqG7iKRsxS1uxl0VPbSo00feiIUdaYslColIM1pn+JHPcIaG2oiF2VWtxq7Mc6fYmh+kGyrcx4x5+gUiP8AQl5LxyTY5708n4AhWWprxDVSkOzKWg0pfkwimbD+BpvsdCXP3WJKj6tBlcOYE6Mqbtkn9vyM5W7jfFTBc1X5MU7DvJgvg9nrCNecNC9CaNJv6Fo0fY57IpKGpEQtTTYFYeaGtavS/RP8QXTUTySGdptdhNwGb3IuuxeKaMNzOzjWPkE/I6lnl+gT8jRn1maXRF0hUQrqTaua+plEUXgE6c1zXuxQramMIuiE6oVxT6kEGEqD+SqJ1ngQmaYJOCGEqlguSX4GqcYTURrOpVlsqr7MhFOaL+DQTql9CHVSnifzneCZnq6SomrY8EkVRYTSRnFlUToEcuDCTCEvXWDi8k9Ju6jxd6PJnEMkN0RzfiSqlLsHzuug+O1NAo23VJmAZVMLKWVWQqpVGs0b5CmfyNG09zZ8PolWq0b/ABVH7gtfEZhEFZa7SmgDgNiZsbFck6UnNJ9/oWjR9jnptGBYktNZLY8jTKcqr0M8P8QXTUneOGlHcblg3aFo+C0/Bt79jJDpxFO7JpFY03H6IHWUcO2JEM9JS6CQlTj1bEN1qcveyZo2ypAKuJBthN9WqLqYIjHC6BjZQYsgJewKql3kgmmrJh6PqQd7lDxOIScG0TsR1CM6TuieX2Yl0RCs1iVR5k+hERZsxiZM0zZIQ0qJXQzFqTVx3QKYoR4NnyPF1u5qk6/5W04dMf0Rr+hUiJo2mUtMmhOhWbbY32+aEHy/SY+lRpNH2Gdt2y/glEvJK/VwM3NSzWv4HkOuj4MHTRMwbvpvAk53XgNvfsbe/YtPwWj4ItLkLV2+haNH2Oe850AHTA964GVEgqqrt/iC6am0CjLdELfk2pvuXPkufJHgm+4mPjX5GFEHoeRzW3dN+RKb9i+hfNewpNpQiCdEJ8Z7riNkyFH6aJpUMXvArYXgYlbQipqY/AHM9SOoXeSrENd+S3grWfwGg9x5cprhple2xK1+SI1UbgkhNpgFpuEGkqaRKegoqM6rJgit4KMEtTCDwpl0JkFNUEKcJiDxjOqllDNPo0yobvgYBPkdAHQ36T95C6CeuyhIaDWDTGuI+Su4hdyJD7p8j9tXz6/hkCdLJuvJgJWkNr9iiiHbmNyXerfklsPyXPksfJrX+pC35IW/JvvYj2f0LRo+xz02aayilEY/8ILpqMhtghppxlG+DfBvg21mxtrNhKqmPfgP2TWHMvdWUijsa0oQTU9aT72LsVaTdRxHUobawMPyZFWJIbBiE/VRCokNqk3yMNkPqMzalsXRhJkoSyprlGOshcZWRYTHKk8oQIVKTNyR6USLzHkhWE1WCyDjWGBNMiirBUp2G7DXbaI8JdP6H739Hjl2KSdiO6M5JokmFdejDHUSvSqVMNuh2S5bRRiJx7Lo+qJrIGt7wn4ka11ioGhhSWmD9ECFe5MPJYeongB+CfFa24jbBNj8Q3QboN8G+BBsSY0WlfoWjR9nnoLpqK+gVqZnOf8AMPTn059OadW4nSE6kNPz0fpjNIjBC/FsR1R4Ll+xER4FLvSLV0m6obFFiPXYRBQS0QTmVHV6CM42mDUSSacXkuhO4ZBkFWjiLZXRjp3gG2Hm3fKJfyCMDZM2cMUxbzWxSG4QhWrHYZ2DSWh0aDIYylSLwH4Dxae7GwO8wooZGiH53NKE1uChJdETKF7vFdxUgQprSb6SOibyUp6VqOpbbJyToruPI9l89r8Bh+PyEmT3E6C9OfTn059OZ8rt/rDBTTalEbR9C0aPs89BdNSfoRpNOFC/vJf3kh/R5v8AiqqqHFexbwGnwC8C0pvQJIVFDkBL8DUS1FOibVR0ImAtR1aPh8jq8OLCI4Wi2QkKjO8jf6rIpqbZnm42kQcMijuYXOqcSTeQjCvLWpiTaz4mmKDrExSY8wUkmKgzssioK/IPK9BiVg/epgl/wzPn2gZR3CdLtH67YMl0ulDrCoQuVC3QnulsQ7222s1VfDUDf5n8BodwJ8Pn/wBGXVVVUv7PMX95L+8l/eSJvRrFrX6Fo0fZ56C6amTXYLrgty78F34LvwXfggwd9i68F14LrwXXgdUDYbwIIK5ufyIh7Cal+RMduJldBhLGiQ6zbFhYn80Q0I4hqo1eiJHgPN8x5qb/ACcoVr5RpwoFlfQiahQzcQROs4xhCZdSRiq0MpT/ALEUNN0ZlnqGjMmzxGycNNMQQbDGtLV1GOhR8XZsSNJ6eKbk/wAjM8cfwCY/A6lJp+4mN2F5TAuvBdeC68F14JMXfYu/Bd+C78F34LvwXfgu/Ahs4w1TF0+haNH2eegumpQ1kQULubu/cTbf7L3OXucvc5c5y5zlznNhfuWH5HSSDVVPiSsgBUJ7vF/gcRpooEkJ26WhJ4KrZi3gr2JG/KghJOKh5DBU0WsCZMFmWQhnLs1inf4HvL4Q2KIdzIXicbFJdHBJCshyeglK+jWZhq2gclL0hNpynDEpB0QasHuISJykT6lFoUJKdJoeENKPDgYUK1V5guPyWD5NlfuWucepv3N3fubu/c3d+5u79zd37m7v3N3fubu/c3d+5V3ITS+z+haNH2eegumovEVKI243j/P23FnX6GTzzzzyrOa2ql7ktpI82K+ROauxjruxvWMKS3dtWOivUJE24KTaJrk1QVKotcSHRNEGKGVFKsgWhbOg2gbY2qQ4tjbEnpUWCQqYmMAs9xVDL/gdWS4NEweCMCNLJkaGmORslRVCYzM6vGz6ZD9vLaP6cMZx27P0JdUm7T/NYf8AV55SlC/4c889r/6//wD/AP8A9iCuGQU7T9C0aPs89BdNRKNpCG1/x/8A/wBhPcD3A9wPcD3A90PZBhWZkSEVsou7GhJWvvZR/slW5zDN6OiJ5pB2JvyaCTF6YIQ2agQJttTIGx2obZ2NyM2faBhWR4SqPP4flC2m7LRjWMWxYRMtrPMiXqOya8G+6RkAOk6eqbqT5u5RMUW7MxfTvuz209wPYD2A9gPYD2A9gPYBL5f/AF//AP8A93J5SSH9C0aPs89BdNRVDSHRM5/8pJJJbMVneCzvBZ3gs7wWd4Le8Dyk9OMoUq0/mbOLhFtSrHNhEqJJvrQqj0xJtoCdWTNPHUKsi0qg0AfDIQ0WCYMoe2BnOicmW6hFf7iCNhR4NYCdDoboc5NThCnfkHoSQhkGQ55qUNUOsFc3RpQJyUqe3GLKX1b4La8FteC3vH10kkkkkkkkkkkkknEtIVXb6Fo0fZ56C6aimXEXKqxRo/8AUpSlKJGA/wCXOVPoHlCNWckY0WSSh+yiwJiHdG6IZlJPRDNqyqzfwKi6hoMo0lMDaaHSOAyCJsHIZFR90Z01MA1IJtiOSEhrJkVu0NmYHq0NCUauth8kS4hkzt/EZqNyVMU0kfkqhaEZfseyLV/39E5z15znJ8WP6KlKUpSlNU72bMhPY/oWjR9nnoLpqOJyEhTnE9v5Pb+T389/Pfz389/PfyDBnVj2A9w/zBVif1BEX+Doju0PopEi30UCFEN4R0kRVjBUXYSGnUT6j1PkvCFTbZvIwRknpuNVshoyNMM58+g3FcoF6RGRRhxFgyC5jTrJDqmjJEeyc+KPYmCrtoV3Th/IOoav8DkmK+jEmKOj8nt/J7fye38nt/J7fye38nt/J7fye38nt/J7fye38nt/J7fye38nt/IskJOZrT6Fo0fZp6HbTIoZpz+ClkTQ0liKTaqm3LOS3PBbngtzwR1m6cBe3gvbwXt4L28F7eCPNv0LD8Fh+Cw/AmOCygX4aK11ijm7hFlioJNsEqupAqzio8tKYNsuguUCxjFCn4RL67CWUazv9Ckpc3Fcw6VOrLMXmQvcHnQ01iQc6vF3Qm3pT2FQuJK06pyJtJIxJ/kdGFsS+KB4DtztN+gm0i3/AIFueC3PBbngtzwW54Lc8FueC3PBbngtzwW54Lc8FueC3PBbngtzwW54Lc8DCYaKkktoX0LRo+0T52xFGNOB7rye68nuvJ7rye68nuvJHj225PZuT2bk9m5PZuT2bk9m5G8taGmJK50aCdCVMSk3q2zGErATMK6o1RNUTe/GbCkzir0ZqioujIfvwNh26RgbPAqrQLVOmDWqKCLkbNNdUiPHZu0OBgkmYRfgeGPQeT3Tk915PdeT3Xk915PdeT3Xk915PdeT3Xk915PdeT3Xk915PdeT3Xk915PdeT3Xk915PdeSJuBvwRjj9C0aPtE+arI4TqX7g1EW2LdwW7gt3BbuC3cFu4Ldwa9zoXjgvHBaOBLdA1f8GBPKoNfIfJSHNbJ+RCEFhbLUjMxnWRgmmqoOQ08yEsJzODGALBW8BznhFWnFDSpQULINWUrcj8ZJtJMMrAqqZdU5GlI2Cya7p+BvwPVfwWrgt3BbuC3cFu4LdwaDLbF+4L9wX7gv3BfuC/cF+4L9wX7gv3BfuC/cF+4L9wX7gWRIilLr9C0aPtE+F4YTShdywvJYXksryX/5L/8AJf8A5L/8l/8AkTcm3UvHyXj5Lx8ls+RnQNfGZGPXEN/AgMFcHDbUeocr8K/x3SibDE53rNkTXwmr8igaMvbMb4sFBlmGuxF75lhENLUCsGJdEpzH8dYz5LK8lleSwvJYXksLyWF5LC8lheSwvJYXksLyWF5LC8lheSwvJYXksLyWF5LC8lheSwvJYXksLySvXDRy+z+haNH2iejRGjqo3g9APQCfH4B6MejCapXwj1J9SW3F/b/jDDDia7F6klnn5DEvuSjzbaNkEslpj0SSyWNHifD3t+jVbOpmITSrZ5CRqWCIfTzajFkBuCDIeaDbbtB2S7WaZ+UKmW0Q+kHoB6AegHoB6AegHoB6AegHoB6AegHoB6AegHoB6AegHoB6AegHpAqQEnop2n6Fo0faJ6dRyQqj1i4PWLg9YuD1i4PWLgbf4cF6uD1q4PWrgvVwetXBergbQ6eScIfpJ7iz+YW9bOntuPaGVv5ZujdG6GoTHQ3oxmTYyHuxNXgx5DuXcMLyaoTGHk2waJNy1i/kRRls2nHZlKXdl+xJ4FweuXB6xcHrFwesXB6xcHrFwesXB6xcHrFwelXB6VcHpVwelXB6VcHpVwelXB6VcHpVwelXB6VcHpVwelXB6VcDGNwIqP6Fo0faJ80RSmLlyXDkuHJcOS4clw5Hlp6D1o9aPWj1g9YEzAt4JBuZMEbBuW6jahvUbi+DcXwbi+DcXwbi+B0M0mMsmOSCrLPa/BKyTDZfhlOdyou5iHsujg+GJISWiWeUELhyXDkuHJcOS4clw5LhyXDkuHJcOSy8lv5LfyW/kt/Jb+S38lv5LfyW/kt/Jb+S38lv5ITijmfQtGj7RPhbwTITa4P/ABDQ7rnwXa4LtcF2uC7XBdfw9M4PROD0Tg9E4MUy1avwLtPCZkTf1OYWrWCHBxwDKNSWiJaIloiWiJaIQ9SaonpqSAonBzWTKFG0P6lKsdJlfDNpUEdU71oNb41wLNi2p4/+hBBBBBBBBBBBBBBBBBBBBBBCS3Buq1ph9C0aPtE97MtGBp/JcOS4clw5NxvuWPksfJY+Sx8ns/J75ye/cialzZTs0I86SbsY6x+5exbEEEEEEECBmQfg9jPDVnwM2hGIVRJ2KE+Kej8lw5LhyXDkuHJcOS4clw5LhyXDkuHJcOS4clw5LhyXDkuHJcOS4clw5LhyXDkuHJcOS4clw5EemVJU18/QtGj7RPWyfUkbfwWfgs/BZ+Cz8EGbfYvHH/Gmml44M90KLwV5Swkj4XBDPB0a+khCEIQhCK+nSrRX3LGSPh8E6nWo/BsN9iz8Fn4LPwWfgs/BZ+Cz8Fn4LPwWfgs/BZ+Cz8Fn4LPwWfgs/BZ+Cz8Fn4LPwWfgs/BZ+Cz8D/ZoQKS+PoWjR/6SfaNH/pJ9o0f+kn2jR/6SfaNH/pJ9o0fYt8EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEFo0f+kf2DR/6R/YNH/pH9g0f/AA//2Q==)
Aprendendo com a luz: é assim que poderia ser a dinâmica de uma onda de luz empregada dentro de uma máquina física de autoaprendizagem. Cruciais são tanto a sua forma irregular como o fato de o seu desenvolvimento ser revertido exactamente a partir do momento da sua maior extensão (vermelho). Crédito: Florian Marquardt, MPL
A inteligência artificial não só proporciona um desempenho impressionante, mas também cria uma procura significativa de energia. Quanto mais exigentes são as tarefas para as quais é treinado, mais energia consome.
Víctor López-Pastor e Florian Marquardt, dois cientistas do Instituto Max Planck para a Ciência da Luz em Erlangen, Alemanha, apresentam um método pelo qual a inteligência artificial poderia ser treinada de forma muito mais eficiente. A sua abordagem baseia-se em processos físicos em vez das redes neurais artificiais digitais utilizadas atualmente. O trabalho está publicado na revista Physical Review X.
A quantidade de energia necessária para treinar o GPT-3, o que torna o ChatGPT um Chatbot eloquente e aparentemente bem informado, não foi revelada pela Open AI, a empresa por trás dessa inteligência artificial (IA). De acordo com a empresa de estatísticas alemã Statista, isto exigiria 1.000 megawatts-hora – quase o mesmo que 200 famílias alemãs com três ou mais pessoas consomem anualmente. Embora este gasto de energia tenha permitido ao GPT-3 aprender se a palavra "profundo" tem maior probabilidade de ser seguida pela palavra "mar" ou "aprendizado" em seus conjuntos de dados , segundo todos os relatos ele não entendeu o significado subjacente de tal frases.
Redes neurais em computadores neuromórficos
A fim de reduzir o consumo de energia dos computadores, e particularmente das aplicações de IA, nos últimos anos várias instituições de investigação têm investigado um conceito inteiramente novo de como os computadores poderão processar dados no futuro. O conceito é conhecido como computação neuromórfica. Embora isto pareça semelhante às redes neurais artificiais, na verdade tem pouco a ver com elas, uma vez que as redes neurais artificiais funcionam em computadores digitais convencionais.
Isto significa que o software, ou mais precisamente o algoritmo, é modelado de acordo com o modo de funcionamento do cérebro, mas os computadores digitais servem como hardware. Eles realizam as etapas de cálculo da rede neuronal em sequência, uma após a outra, diferenciando entre processador e memória.
“A transferência de dados entre esses dois componentes por si só consome grandes quantidades de energia quando uma rede neural treina centenas de bilhões de parâmetros, ou seja, sinapses, com até um terabyte de dados”, diz Marquardt, diretor do Instituto Max Planck para a Ciência. da Luz e professor da Universidade de Erlangen.
O cérebro humano é totalmente diferente e provavelmente nunca teria sido evolutivamente competitivo se tivesse funcionado com uma eficiência energética semelhante à dos computadores com transistores de silício. Provavelmente teria falhado devido ao superaquecimento.
O cérebro é caracterizado por realizar as inúmeras etapas de um processo de pensamento em paralelo e não sequencialmente. As células nervosas, ou mais precisamente as sinapses, são processador e memória combinados. Vários sistemas em todo o mundo estão sendo tratados como possíveis candidatos às contrapartes neuromórficas das nossas células nervosas , incluindo circuitos fotônicos que utilizam luz em vez de elétrons para realizar cálculos. Seus componentes servem simultaneamente como interruptores e células de memória.
![](data:image/jpeg;base64,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)
Inteligência artificial como uma fusão de pinball e ábaco: neste experimento mental, o
pinball azul com carga positiva representa um conjunto de dados de treinamento. A bola
é lançada de um lado a outro do prato. Crédito: Florian Marquardt, MPL
Uma máquina física de autoaprendizagem otimiza suas sinapses de forma independente
Juntamente com López-Pastor, estudante de doutorado no Instituto Max Planck para a Ciência da Luz, Marquardt desenvolveu agora um método de treinamento eficiente para computadores neuromórficos. “Desenvolvemos o conceito de máquina física de autoaprendizagem”, explica Florian Marquardt. “A ideia central é realizar o treinamento na forma de um processo físico, em que os parâmetros da máquina são otimizados pelo próprio processo”.
Ao treinar redes neurais artificiais convencionais, é necessário feedback externo para ajustar a força dos muitos bilhões de conexões sinápticas. “Não exigir esse feedback torna o treinamento muito mais eficiente”, diz Marquardt. Implementar e treinar uma inteligência artificial em uma máquina física de autoaprendizagem não só economizaria energia, mas também tempo de computação.
“Nosso método funciona independentemente de qual processo físico ocorre na máquina de autoaprendizagem, e nem precisamos saber o processo exato”, explica Marquardt. "No entanto, o processo deve cumprir algumas condições. Mais importante ainda, deve ser reversível, o que significa que deve ser capaz de avançar ou retroceder com um mínimo de perda de energia."
“Além disso, o processo físico deve ser não linear, ou seja, suficientemente complexo”, diz Marquardt. Somente processos não lineares podem realizar as complicadas transformações entre dados de entrada e resultados. Uma bola de pinball rolando sobre um prato sem colidir com outro é uma ação linear. Porém, se for perturbado por outro, a situação torna-se não linear.
Teste prático em um computador neuromórfico óptico
Exemplos de processos reversíveis e não lineares podem ser encontrados na óptica. Na verdade, López-Pastor e Marquardt já estão colaborando com uma equipe experimental no desenvolvimento de um computador neuromórfico óptico. Esta máquina processa informações na forma de ondas de luz sobrepostas, através das quais componentes adequados regulam o tipo e a força da interação. O objetivo dos pesquisadores é colocar em prática o conceito de máquina física de autoaprendizagem.
“Esperamos poder apresentar a primeira máquina física de autoaprendizagem em três anos”, diz Florian Marquardt. Até então, deverá haver redes neurais que pensem com muito mais sinapses e sejam treinadas com quantidades de dados significativamente maiores do que as atuais.
Como consequência, provavelmente haverá um desejo ainda maior de implementar redes neurais fora dos computadores digitais convencionais e de substituí-las por computadores neuromórficos treinados de forma eficiente. “Estamos, portanto, confiantes de que as máquinas físicas de autoaprendizagem têm grandes chances de serem utilizadas no desenvolvimento da inteligência artificial”, afirma o físico.
Mais informações: Víctor López-Pastor et al, Máquinas de autoaprendizagem baseadas em retropropagação de eco hamiltoniano, Physical Review X (2023). DOI: 10.1103/PhysRevX.13.031020
Informações do periódico: Revisão Física X